Endodermal Germ-Layer Formation Through Active Actin-Driven Migration Triggered by N-Cadherin

Endodermal germ-layer formation through active actin-driven migration triggered by N-cadherin

Florence A. Gigera,b,c and Nicolas B. Davida,b,c,d,1

aCNRS UMR8197, F-75005 Paris, France; bINSERM U1024, F-75005 Paris, France; cInstitut de Biologie de l’Ecole Normale Supérieure, F-75005 Paris, France; and dLaboratory for Optics and Biosciences, Ecole Polytechnique, 91128 Palaiseau, France

Edited by Roeland Nusse, Stanford University School of Medicine, Stanford, CA, and approved August 16, 2017 (received for review May 16, 2017) Germ-layer formation during gastrulation is both a fundamental We show that cell internalization relies on an active, actin-driven step of development and a paradigm for tissue formation and migration process. Rather than being attracted to their destina- remodeling. However, the cellular and molecular basis of germ- tion, cells migrate away from their neighbors in a process medi- layer segregation is poorly understood, mostly because of the lack ated by Rac1 and triggered by cadherin-2 (hereafter referred to of direct in vivo observations. We used mosaic zebrafish embryos as “N-cadherin”). to investigate the formation of the endoderm. High-resolution live imaging and functional analyses revealed that endodermal cells Results reach their characteristic innermost position through an active, Endodermal Cells Emit Cytoplasmic Extensions Toward the Yolk oriented, and actin-based migration dependent on Rac1, which con- Syncytial Layer and Rapidly Migrate to Its Surface. To unravel the trasts with the previously proposed differential adhesion cell sorting. basis for germ-layer formation, we analyzed the cell behavior and Rather than being attracted to their destination, the yolk syncytial dynamics of endodermal cells during internalization. Naive cells layer, cells appear to migrate away from their neighbors. This migra- can easily be driven to adopt an endodermal fate (SI Experi- tion depends on N-cadherin that, when imposed in ectodermal cells, is mental Procedures and refs. 5, 6, and 11). Combined with cell sufficient to trigger their internalization without affecting their fate. transplants, this allows the creation of mosaic embryos, a pre- Overall, these results lead to a model of germ-layer formation in requisite to good imaging. At the late blastula stage, single en- which, upon N-cadherin expression, endodermal cells actively mi- dodermal progenitors expressing the actin-labeling construct grate away from their epiblastic neighbors to reach their internal Lifeact-GFP were transplanted close to the margin of embryos position, revealing cell-contact avoidance as an unexplored mech- expressing membrane-bound mCherry. Rapid 4D confocal im- anism driving germ-layer formation. aging was used to acquire entire volumes over time, and optical sections were reconstructed to analyze cell behavior in the plane gastrulation | endoderm | cell migration | zebrafish | cadherin of the internalization movement (Fig. 1A). When in the epiblast, endodermal cells emitted large, frequent, astrulation is the first developmental stage when different and short-lived cytoplasmic extensions enriched in actin and ori- Gprogenitors segregate and organize into distinct germ layers. ented toward the yolk syncytial layer (YSL) (n = 103 extensions, Large-scale cell movements set up the body plan of the embryo, Puniform/measured angle distribution < 0.001) (Fig. 1C). Cells internalized with endodermal and mesodermal cells becoming internalized through rapid migration to the surface of the YSL (mean speed: − beneath the ectoderm. Whereas an extensive body of literature 2.4 μm·min 1; n = 6cells)(Fig.1B, D,andE and Movie S1). They describes the pathways that specify endoderm and mesoderm later differentiated into endodermal derivatives (Fig. 1F and ref. identity (1), the cellular mechanisms that physically create these 6). To rule out artifacts due to cell transplants or endoderm in- layers in the embryo are much less understood. duction, we used mosaic expression of Lifeact-GFP in wild-type In frog, internalization is achieved by involution, in which the embryos to look at the behavior of endogenous untreated endo- prospective mesoderm and endoderm roll inward as a coherent derm. We focused on cells located in the four most marginal rows, tissue at the blastopore, driven by the vegetal rotation of the endodermal mass (2). In amniotes, as well as in the sea urchin and Significance urodeles, internalization is achieved via ingression of single cells, with endodermal and mesodermal progenitors undergoing an Construction of the adult body during development implies the epithelial-to-mesenchymal transition (EMT) (3, 4). In fish, cell separation of cells into tissues and organs. The first of these transplantation experiments have demonstrated that cells inter- events occurs during gastrulation, when cells segregate in germ nalize individually but in a coordinated manner, termed “syn- ” – layers, setting the bases of the body plan. Despite its impor- chronized ingression, at the very margin of the blastoderm (5 7). tance, the molecular and cellular basis of germ-layer formation The general movements leading to germ-layer formation have has remained elusive. This work uses live imaging and functional thus been described in many species. However, how these move- approaches to reveal how the endodermal layer forms in ments are driven at the cellular scale remains poorly understood zebrafish, leading to an original model of cell segregation by (4). Sixty years ago, Townes and Holtfreter established that, when active migration away from neighboring cells, a process trig- dissociated and mixed, embryonic cells would sort into their pre- gered by N-cadherin. We propose this may be a core conserved viously specified germ layers (8). Following this original observa-

mechanism driving germ-layer formation and might be ex- BIOLOGY tion, it was proposed that the mechanism underpinning germ-layer tended to other processes of cell segregation, highlighting gas- DEVELOPMENTAL formation in vivo could be cell sorting based on differential ad- trulation as a paradigm for tissue formation. hesion (9, 10) and/or differential cortical tension (11). However, this hypothesis has not been validated in the embryo (12). Author contributions: F.A.G. and N.B.D. designed research, performed research, analyzed Our limited understanding of the mechanisms driving cell in- data, and wrote the paper. ternalization and germ-layer formation most likely stems from the The authors declare no conflict of interest. difficulty and hence the limited number of direct in vivo obser- This article is a PNAS Direct Submission. vations, in particular in vertebrates (4, 13, 14). Here, we focused 1To whom correspondence should be addressed. Email: [email protected]. on endodermal cells in the zebrafish embryo to decipher the This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. molecular and cellular mechanisms driving germ-layer separation. 1073/pnas.1708116114/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1708116114 PNAS | September 19, 2017 | vol. 114 | no. 38 | 10143–10148 Downloaded by guest on September 29, 2021 Fig. 1. Endodermal cells rapidly migrate to the YSL. (A) Lifeact-GFP–expressing endodermal cells were transplanted to the margin of embryos expressing membrane-bound mCherry, just beneath the EVL. 4D stacks were acquired, and sagittal sections were reconstructed. (B) Sagittal sections showing endodermal cell internalization (Movie S1). Arrowheads point to actin-rich cytoplasmic extensions. The dashed line delineates the surface of the embryo; the dotted line represents the limit between the blastoderm and the YSL. Animal pole is to the top. (C)Orientationof actin-rich protrusions. Dots represent measurements. (D) Movement of internalizing cells along the surface– YSL axis. The spatial and temporal origin is defined as the beginning of migration. (E) Instant speed of internalizing cells along the surface–YSL axis. Mean speed for each cell is represented as a colored seg- ment. (F) Transplanted endodermal cells contribute to endodermal derivatives at 24 hpf (arrowhead). (Scale bar: 20 μm.)

which contain endodermal precursors (15). These cells exhibited cells into their animal pole (Fig. 2C). We checked that transplanted the same behavior as transplanted ones, extending actin-rich cyto- and surrounding cells expressed the endodermal marker sox32 (Fig. plasmic extensions toward the YSL and migrating to its surface 2 H and I) and displayed similar levels of cadherin-1 (hereafter re- − (mean speed: 1.7 μm·min 1; n = 7cells)(Movie S2). ferred to as “E-cadherin”)(Fig.2J and K), which accounts for about We noticed that endodermal cells internalized either inde- 80% of cell adhesion at this stage (11). Contrary to ectodermal cells pendently of neighboring cells or in coordination with them (Fig. that stay at the surface when surrounded by identical neighbors, S1 A, A′, B, and B′ and compare Movies S3 and S4), which is about half of the transplanted endodermal cells internalized (n = consistent with previous reports showing that internalization of 21 embryos) (Fig. 2 F and G). Live analysis of these internalizing hypoblastic cells is a more coherent process at the ventral than at cells revealed that they emitted actin-rich protrusions toward the the dorsal margin (16). Coordinated internalization likely cor- YSL, as observed at the margin of the embryo (n = 6 cells for each relates with nonautonomous effects that were first identified condition) (Movies S7 and S8). As could be expected, we noticed using maternal and zygotic one-eyed pinhead (MZoep) mutant that when all cells are endodermal, they all tend to internalize. embryos that are deficient in Nodal signaling and do not gas- However, because of steric constraints, only some of them can reach trulate: When transplanted to the margin of a wild-type embryo, the surface of the YSL (Movie S8), which likely explains why only a cell from an MZoep mutant embryo is driven into the hypoblast part of the transplanted cells internalize in this condition. by its neighbors (Fig. S1C, Movie S5, and ref. 5). The movement To further test if cells can internalize irrespective of differential of a cell may thus result both from its own activity and from the adhesion, we artificially lowered the adhesive properties of in- behavior of its neighbors. To circumvent these nonautonomous ternalizing cells or of surrounding cells. Since at this stage E-cadherin effects, we analyzed the behavior of wild-type endodermal cells is responsible for 80% of cell adhesion, we down-regulated transplanted to the margin of nongastrulating MZoep embryos. E-cadherin by morpholino (Fig. S2A) in either the internalizing Even though neighboring cells did not internalize, transplanted endodermal cells or their ectodermal neighbors. Neither of these cells internalized with the same internalization features as in affected endoderm internalization (Fig. S2B), demonstrating that wild-type embryos (n = 3 cells) (Fig. S1D and Movie S6), demon- differences in E-cadherin–based adhesion are not driving endo- strating a cell-autonomous process. derm internalization. Consistently, we observed similar E-cadherin In summary, endodermal cells internalize autonomously by expression levels in ectodermal and endodermal cells (Fig. S2C). emitting long, actin-rich cytoplasmic extensions directed toward Together, these results demonstrate that reducing or reversing the YSL and by rapidly migrating to the surface of the YSL. differential adhesion does not prevent endoderm internalization, which can be achieved by active cell migration. Active Migration Is Sufficient to Ensure Endoderm Internalization. Although it has been proposed that differences in cortical tension The Internalization of Endodermal Cells Is Dependent on Rac1 and and adhesive properties between germ-layer progenitors constitute Arp2/3. As the internalization of endodermal cells appeared to a key factor triggering their segregation (11), our observations of be an active process, we tested the potential role of the small oriented actin-rich protrusions and rapid internalization strongly GTPases RhoA, Cdc42, and Rac1, which are established regu- suggest that endodermal cells can internalize by active migration. lators of cell migration (17). To do so, we interfered with the To directly test this, we analyzed if endodermal cells can internalize function of each protein in turn and analyzed the internalization when differences in adhesion with their neighbors are abolished of endodermal cells transplanted to the animal pole of wild-type or reversed. embryos (Fig. 3A) rather than at the margin, since at the margin We first modulated differential adhesion by modifying the identity even noningressing cells can be internalized by their neighbors. of neighboring cells. As controls, we transplanted either ectodermal Expression of dominant negative forms of RhoA or Cdc42 did or endodermal cells into the outermost layer of the ectoderm (ani- not affect cell internalization (ncontrol = 60 embryos; ndnRhoA = mal pole) of a host embryo (Fig. 2 A and B). As expected, when 45 embryos, P = 0.43; ndnCdc42 = 31 embryos, P = 0.46) (Fig. 3B). surrounded by identical cells, ectodermal cells remained at their However, endodermal cells expressing a dominant negative form position (n = 20 embryos) (Fig. 2 D and G), while endodermal cells of Rac1 did not internalize (ndnRac1 = 26 embryos, P < 0.001) surrounded by ectodermal cells internalized during gastrulation (n = (Fig. 3B) or contribute to endodermal derivatives at 24 h post 17 embryos) (Fig. 2 E and G). We then induced whole embryos to fertilization (hpf) (ncontrol = 51 embryos, ndnRac1 = 28 embryos, form endoderm and transplanted differently labeled endodermal Pcontrol/dnRac1 < 0.001) (Fig. S3 A, B, and D). Importantly, when

10144 | www.pnas.org/cgi/doi/10.1073/pnas.1708116114 Giger and David Downloaded by guest on September 29, 2021 Fig. 2. Active migration is sufficient to ensure endoderm internalization. (A–F) Ectodermal cells (A and D) or endodermal cells (B, C, E,andF) were transplanted to the superficial-most layer of a blastoderm of either ectodermal (A, B, D,andE) or endodermal (C and F) identity. Sections show the position of transplanted cells at the onset of gastrulation (A–C) and at midgastrulation (D–F). (G) Distribution of embryos according to the percentage of internalized cells at midgastrulation plotted as a cumulative plot. ***P < 0.001. (H–K) Endodermal cells (red nuclear labeling) were transplanted into ectodermal (H and J) or endodermal (I and K) neighboring cells. (H and I) Expression of the endodermal marker sox32 revealed by in situ hybridization. (J and K) E-cadherin immunohistochemistry revealed

similar protein levels at the interface between two transplanted cells (red nuclear labeling) and between two host neighboring cells (ntransplanted endo > ecto = 11, necto = 13, Ptransplanted endo/ecto = 0.83, ntransplanted endo > endo = 10, nendo = 10, Ptransplanted endo/endo = 0.69). (Scale bars: 20 μm.)

transplanted to a deep position in the embryo, at the surface of the actin branching complex Arp2/3, a direct regulator of protrusions YSL, most endodermal cells expressing dominant negative Rac1 (18). Endodermal cells expressing the VCA fragment internalized n = P < B did form endodermal derivatives at 24 hpf, showing that Rac1 is poorly ( VCA 22 embryos, 0.001) (Fig. 3 ), consistent with a required specifically for the internalization step (ndnRac1 deep = 37 role of actin-rich protrusions in cell internalization. Interestingly, while ectodermal cells do not internalize, they embryos, PdnRac1 surface/deep < 0.001) (Fig. S3 C and D). To unravel the role of Rac1 in the internalization process, we display oriented protrusions similar to those observed in endo- observed the morphology and dynamics of endodermal cells dermal cells (Figs. S4 and S5 and Movie S10). This demonstrates expressing dominant negative Rac1. These cells showed a dra- that, in addition to the acquisition of polarization and protrusive matic reduction in the frequency of cytoplasmic extensions activity dependent on Rac1 and Arp2/3, another mechanism is −1 required to trigger endoderm internalization. (frequencycontrol = 1.4 extensions·min , n = 88 extensions; fre- −1 quencydnRac1 = 0.3 extensions·min , n = 57 extensions; P < 0.01) G Endodermal Cells Are Not Attracted to the YSL but Migrate Away (Fig. 3 and Movie S9). The remaining extensions were shorter from Their Neighbors. Because cells are polarized toward the YSL = μ = μ P < C (lengthcontrol 12.4 m, lengthdnRac1 7.6 m, 0.01) (Fig. 3 , and migrate to its surface, the YSL, which is transcriptionally ac- E H P = ,and ) but still polarized toward the YSL ( angle control/dnRac1 tive, could be the source of an attractant. To test this, we injected 0.3) (Fig. 3 D and F). These results suggest that cytoplasmic ex- RNase with rhodamine-dextran into the YSL at the midblastula tensions are required for cell internalization. To confirm this, stage. RNase injection into the YSL has been shown to eliminate we used the constitutively active verprolin, cofilin, acidic (VCA) RNAs rapidly (19), which we verified by in situ hybridization for domain of neural (N)-WASP to interfere with the function of the cb120 and A2ML that are normally strongly expressed in the YSL

Internalised cell A C Surface D Surface Surface

Fig. 3. Rac1 and Arp2/3 are required for the in- Control Yolk ternalization of endodermal cells. (A) Endodermal Non-internalised cell Yolk cells were transplanted into the superficial-most Surface Lifeact-GFP Mb mCherry YSL layer of the blastoderm. By midgastrulation, cells in E F contact with the surface of the YSL were counted as Surface Late blastula Mid-gastrula internalized, and cells in contact with the EVL were counted as noninternalized. (B) Distribution of em- Yolk bryos according to the percentage of internalized cells at midgastrulation plotted as a cumulative plot. + dnRac1

While most control, dnRhoA-, and dnCdc42-expressing BIOLOGY B 1 endodermal cells were internalized, endodermal Lifeact-GFP cells expressing either dnRac1 or VCA were not. (C DEVELOPMENTAL 0.8 Mb mCherry YSL and E) Cytoplasmic extensions (arrowheads) emitted G 3 H 30 ) C E 0.6 -1 ** ** by control ( ) or dnRac1-expressing ( ) endodermal 25 cells labeled with Lifeact-GFP. (D and F) Orientation Control 0.4 ns 2 20 D dnRhoA ns of cytoplasmic extensions emitted by control ( )or *** dnCdc42 *** 15 dnRac1-expressing (F) endodermal cells. (G and H) 0.2 dnRac1 G H VCA 1 10 Frequency ( ) and length ( ) of cytoplasmic exten-

Cumulative proportion of embryos 0 sions emitted by control or dnRac1-expressing en- Length of extensions (µm) 5 μ 08604020 0100 Frequency of extensions (min dodermal cells. (Scale bars: 10 m.) ns, nonsignificant 0 0 Percentage of internalised cells Control dnRac1 Control dnRac1 (P > 0.05); **P < 0.01; ***P < 0.001.

Giger and David PNAS | September 19, 2017 | vol. 114 | no. 38 | 10145 Downloaded by guest on September 29, 2021 (Fig. S6 C–F and refs. 20 and 21). RNA depletion led to severe (Fig. 5A). This could be rescued by coinjection of an N-cadherin mRNA B defects, with the blastoderm lifting off the yolk cell (Fig. S6 and insensitive to the morpholino (n = 69 embryos, Pcontrol/N-cad rescue = ref. 19). Nevertheless, endodermal cells transplanted beneath the 0.14) (Fig. 5A). To confirm the role of N-cadherin in cell inter- enveloping layer (EVL) of RNase-injected embryos (Fig. S6 A nalization, we transplanted n-cadherin (pac) mutant endodermal and G) had reached the surface of the YSL by midgastrulation cells at the margin of MZoep embryos (which do not gastrulate). n = n = (assessed on control embryos, RNase 9/10 embryos, control N-cadherin loss of function significantly impaired endodermal 14/14 embryos, P = 0.4) (Fig. S6H). This demonstrates either cell internalization (nsiblings = 62 embryos, npac−/− = 33 embryos, that the internalization of endodermal cells is independent of the P < 0.05) (Fig. S8), confirming the role of N-cadherin in the in- YSL or that the internalizing signal is provided by proteins ternalization process. produced before RNA depletion. Live imaging revealed that cells with reduced N-cadherin To discriminate between these possibilities, we examined en- levels produced cytoplasmic extensions directed toward the doderm cell behavior in the absence of a YSL. We transplanted YSL at the same length and frequency as control endodermal cells cells to the animal pole of host embryos at the late blastula stage (Pangle control/Mo N-cad = 0.96, lengthcontrol = 12.4 μm, lengthMo N-cad = and dissected animal cap explants at the onset of gastrulation to μ P = = · −1 A 11.8 m, control/Mo N-cad 0.95; frequencycontrol 1.47 extensions min , separate the blastoderm from the YSL (Fig. 4 ). The EVL also −1 frequencyMo N-cad = 1.06 extensions·min , Pcontrol/Mo N-cad = 0.15) detached from these explants. After 3 h of culture, when non- C E F – dissected embryos had reached midgastrulation, transplanted con- (Fig. 5 , ,and and Movie S13). N-cadherin defective endo- trol ectodermal cells were still within the explant (n = 11/11 dermal cells thus resembled ectodermal cells, which emit long, explants) (Fig. 4B). In contrast, endodermal cells were found out- actin-rich cytoplasmic extensions toward the YSL but do not mi- side the blastoderm explant (n = 12/16 explants, Pecto/endo < 0.001) grate to the YSL (Fig. S4 and Movie S10). This suggested that (Fig. 4C). Live analysis immediately after dissection revealed N-cadherin is not required for the initial step of cell polarization that endodermal cells transplanted into the outermost layer of but may be responsible for triggering the specific migration behavior cells emitted large, actin-rich cytoplasmic extensions directed of endodermal cells. We directly tested this by misexpressing N- outwards and migrated out of the explant (n = 15 cells) (Fig. 4D cadherin in ectodermal cells. Strikingly, expression of N-cadherin and Movie S11). Endoderm migration out of the epiblast is was sufficient to induce internalization of ectodermal cells (ncontrol = thus YSL independent, suggesting that endodermal cells in- 43 embryos, nN-cad = 62 embryos, P < 0.001) (Fig. 5 B and D–F and ternalize by migrating away from their neighbors to reach a cell- Movie S14) without inducing an endodermal identity (Fig. S7 I– free area. L). N-cadherin is thus sufficient to trigger cell-autonomous cell Consistent with this idea, we observed instances in which an ingression without affecting cell fate. To ensure that this inter- enveloping layer had partially reformed around the explant leav- nalization is not due to cell sorting by differential adhesion, we took ing only a few gaps through which blastoderm cells were in direct advantage of a truncated form of N-cadherin lacking the cytoplasmic contact with the medium. Strikingly, when an endodermal cell domain (N-cadΔcyto), which cannot bind the cortical cytoskeleton produced a protrusion reaching the gap, the cell would rapidly n = and is therefore unable to transmit adhesion forces (26, 27). Re- migrate to and then through the gap, exiting the explant ( markably, misexpression of this truncated form of N-cadherin also 3cells)(Movie S12). triggered ectodermal cell internalization (n = 41 embryos, P < B Role of N-Cadherin in Triggering Cell Internalization. We looked for 0.001) (Fig. 5 ). pathways that could induce this endoderm-specific behavior. Discussion Interestingly, in many species N-cadherin expression is turned on at the onset of gastrulation in cells that will internalize (14, 22, In this study, we have analyzed the mechanisms driving germ- 23) and is known to induce cell migration of some cell types (24). layer formation during gastrulation. Our results lead to a two- In fish, N-cadherin expression is specifically turned on at the step model for endoderm internalization. Before the onset of onset of gastrulation in internalizing cells (25), in response to gastrulation, cells, regardless of their identity, are polarized and Nodal signaling (Fig. S7 A–H). We thus tested if N-cadherin emit actin-rich, Rac1-dependent protrusions to the surface of the could be involved in endodermal cell internalization. YSL. We hypothesize that these protrusions allow endodermal Inhibition of N-cadherin by morpholino impaired the internali- cells to identify the YSL as a region free of neighboring cells. At zation of endodermal cells transplanted to the animal pole of host the onset of gastrulation, the expression of N-cadherin, induced by embryos (ncontrol = 88 embryos, nMo N-cad = 205 embryos, P < 0.001) Nodal signaling, triggers endoderm internalization. Cells actively

Fig. 4. Endodermal cells are not attracted to the YSL. (A) Ectodermal or endodermal cells were transplanted to the animal pole of a host embryo. An animal cap explant was then dissected and cultured. (B and C) After 3 h of culture, transplanted ectodermal cells are still in the cap explant (B), while endodermal cells have exited (C). (B and C, Upper)Topview.(B and C, Lower) Section along the dotted line in the upper view; brackets show the position of the animal cap. (D) Endodermal cells migrate out of the explant (Movie S11) using actin-rich protrusions pointing outwards (arrowheads). (Scale bars: 10 μm.)

10146 | www.pnas.org/cgi/doi/10.1073/pnas.1708116114 Giger and David Downloaded by guest on September 29, 2021 Fig. 5. Role of N-cadherin in inducing cell internalization. (A) Control endodermal cells or endodermal cells injected with N-cadherin morpholino or with N-cadherin morpholino and N-cadherin mRNA were transplanted to the animal pole of a host embryo. At midgastrulation, N-cadherin morphant cells had internalized poorly, which can be rescued by expression of a morpholino-insensitive N-cadherin RNA. Distribution of embryos according to the percentage of internalized cells at midgastrulation is plotted as a cumulative plot. (B) Control ectodermal cells or ectodermal cells injected with N-cadherin mRNA (full length or lacking the cytoplasmic part) were transplanted to the animal pole of a host embryo. Contrary to control ectodermal cells, at midgastrulation, N-cadherin–expressing ectodermal cells had internalized. (C) Orientation of protrusions emitted by N-cadherin morphant endodermal cells. (D) Orientation of protrusions emitted by N-cadherin–expressing ectodermal cells. (E and F) Frequency (E) and length (F) of cytoplasmic extensions emitted by N-cadherin morphant endodermal cells and N-cadherin–expressing ectodermal cells compared with control

endodermal and ectodermal cells. ncontrol endo = 81, nendo Mo N-cad = 108, ncontrol ecto = 102, necto N-cad = 99 extensions. ns, nonsignificant (P > 0.05); ***P < 0.001.

migrate away from their neighbors to adopt their characteristic have also been described in ingressing cells in mouse, with Rac1- innermost position at the surface of the YSL. deficient cells losing the ability to migrate away from the primitive One of the main endeavors of this work was to provide high- streak (30). In addition, N-cadherin is also expressed at the onset of resolution live observations of ingressing cells and combine them gastrulation in cells that will internalize in chick and mouse, as well with functional analyses to directly assess the mechanisms of as in fly (14, 22, 23). In these species, N-cadherin up-regulation is endoderm formation in vivo. Thus far, unraveling these mecha- accompaniedbyE-cadherindown-regulation.Thisswitchincadherin nisms has been largely hindered by the complexity of the system: is classically interpreted as a mechanism inducing cell sorting In particular, endoderm internalization is accompanied by the through differential adhesion (10). However, our results show massive internalization of mesodermal cells, which is sufficient to that, in fish, cells instead internalize through active migration, and, drive noningressing cells into the hypoblast (Fig. S1C, and Movie accordingly, there is no down-regulation of E-cadherin in inter- S5) (5). This community effect, as well as the potential existence nalizing cells (Fig. S2 and refs. 11 and 31). This either suggests a of redundant pathways, likely explains why genetic screens have fundamental difference in internalization mechanisms between largely failed to identify genes involved in the internalization pro- fish and other species or implies that the role of N-cadherin in cess. By circumventing these non–cell-autonomous effects, we have other species should be reconsidered. In support of the latter idea, revealed some of the cellular machinery controlling endoderm internalization was recently shown to be independent of E-cadherin formation. Specifically, while no gastrulation defect has been ob- down-regulation in both fly and chick (32, 33). In different epi- served in n-cadherin mutant embryos (28), experiments using cell thelial cancer cell lines, N-cadherin can similarly promote EMT transplants revealed a role for N-cadherin in controlling inter- without the loss of E-cadherin (34). In these systems, N-cadherin nalization behavior without affecting cell fate. Further work will functions by promoting cell migration independently of cell ad- be required to identify the pathway acting downstream of N-cadherin hesion. In breast cancer cell lines, for instance, the extracellular in endodermal cells. domain of N-cadherin was shown to interact with FGF receptor Our live observations, showing oriented actin-rich protrusions 1, prevent its internalization, and lead to sustained FGF signal- and rapid cell translocation, as well as direct modulation of the ing, which promotes migration (for a review, see ref. 35). cell environment and experiments using the extracellular domain We propose that at the onset of gastrulation N-cadherin up- of N-cadherin, demonstrated that endodermal cells internalize by

regulation induces cell motility and allows internalizing cells to BIOLOGY an active migration and did not support a key role for a cell sorting actively migrate away from their neighbors. This would constitute DEVELOPMENTAL based on differential adhesion, as previously proposed (8, 11). a core mechanism driving germ-layer formation, which in some Xenopus This is in direct agreement with studies performed in , species is accompanied by a loss of E-cadherin expression, prob- showing that differential adhesion leads to cell sorting in vitro but ably depending on the degree of epithelization of the epiblast at not in vivo (12). However, it is possible that differential adhesion the onset of gastrulation (3). acts redundantly or is involved in the maintenance of germ-layer separation, as suggested by modeling (29). Experimental Procedures The involvement of active migration in endoderm formation in Embryos. Embryos were obtained by natural spawning of wild-type, oeptz57−/− fish suggests a possible conservation of gastrulation mechanisms (36), or pactm101+/− (28) adult fish. All animal studies were done in accordance between fish and amniotes. Indeed, Rac1-dependent cell protrusions with the guidelines issued by the French Ministry of Agriculture and were

Giger and David PNAS | September 19, 2017 | vol. 114 | no. 38 | 10147 Downloaded by guest on September 29, 2021 approved by the Direction Départementale des Services Vétérinaires de Paris Time-Lapse Imaging. Dechorionated embryos were mounted in 0.2% agarose in and the Charles Darwin Ethical Committee (C2EA-05). embryo medium and imaged from the germ-ring stage (5.7 hpf) to the 70% epiboly stage (8 hpf). Imaging was performed on an inverted thermostated Cell Transplantation Experiments. At the 30% epiboly stage (4.7 hpf), Nikon spinning-disk microscope (Yokogawa) equipped with an Evolve camera 5–20 cells from donor embryos were transplanted to the margin or animal (Photometrics) and MetaMorph software (Molecular Devices). Z-stacks with a pole of host embryos (37). Embryos were then cultured in embryo medium Z-step of 1 μm were collected at 1-min intervals, and images were recon- μ (38) with 10 U/mL penicillin and 10 g/mL streptomycin (Thermo Fisher). For structed using Fiji software. single-cell transplantations, donor embryos were cultured in calcium-free Endodermal cells were automatically tracked using Imaris software (Bitplane). Ringer medium after dechorionation at the sphere stage (4 hpf). At 30% Tracks were manually validated and corrected when necessary. Neighbor cells epiboly, cells were mechanically dissociated, and isolated cells were trans- tm101 labeled with membrane-bound mCherry were tracked manually using a custom- planted. For experiments with n-cadherin pac mutant embryos, donor modified version of the manual tracking Fiji plugin. Data were further processed embryos were genotyped after transplantation (primers for amplification of the genomic region to sequence were GTGTCTGCTTGTGTGTTCTTG and with MATLAB. TGATCGCCTGACAACACTATT). ACKNOWLEDGMENTS. We thank A. Bonnet, J. Clarke, C. Houart, and F. Rosa for Injections into the YSL. Rhodamine-dextran 10,000 MW (2 mg/mL; Thermo Fisher critically reading the manuscript; L. Bally-Cuif, M. Hammerschmidt, C.-P. Heisenberg, and C. Revenu for mutants, plasmids, and antibodies; Firas Bouallague for fish care; D1824) was injected alone or in combination with DNase-free RNase (0.5 mg/mL E. Hirsinger, F. Giudicelli, and the Institute of Biology Paris-Seine Fish Facility; and diluted to 10 μg/mL; Roche) into the YSL of high-stage embryos (3.3 hpf). theInstitutdeBiologiedel’Ecole Normale Supérieure Imaging Facility, which received support from the Region Ile de France (Grants NERF 2009 44 and NERF 2011 Culture of Animal Caps. Cells were transplanted to the animal pole at the 30% 45), the Fondation pour la Recherche Médicale (Grant FRM DGE 20111123023), epiboly stage. At 50% epiboly (5.3 hpf), animal caps were dissected with fine andtheAgenceNationaledelaRecherche(Grants ANR-10-INSB-04-01, ANR-10- tweezers. The EVL detached from the cap. The blastoderm explant was in- LABX-54-MEMO-LIFE, and ANR-11-IDEX-0001-02-PSL*). This work was supported cluded and cultured in 0.2% agarose in L15 (65%; Gibco), embryo medium by Fondation ARC pour la Recherche sur le Cancer Grants PJA 20131200143 and (20%), BSA (1 mg/mL; Eurobio), Hepes (pH 7.5, 10 mM; Sigma), penicillin (10 U/mL), PJA 20151203256. F.A.G. was supported by the Ministère de l’Enseignement and streptomycin (10 μg/mL; Thermo Fisher). For live imaging, explants were Supérieur et de la Recherche and Fondation pour la Recherche Médicale Grant imaged immediately. The position of transplanted cells was assessed when FDT20150532614. N.B.D. was supported by the Centre National de la Recherche control nondissected embryos had reached 70% epiboly (8 hpf). Scientifique.

1. Schier AF, Talbot WS (2005) Molecular genetics of axis formation in zebrafish. Annu 25. Warga RM, Kane DA (2007) A role for N-cadherin in mesodermal morphogenesis Rev Genet 39:561–613. during gastrulation. Dev Biol 310:211–225. 2. Winklbauer R, Schürfeld M (1999) Vegetal rotation, a new gastrulation movement in- 26. Maître J-L, et al. (2012) Adhesion functions in cell sorting by mechanically coupling volved in the internalization of the mesoderm and endoderm in Xenopus. Development the cortices of adhering cells. Science 338:253–256. 126:3703–3713. 27. Chu YS, et al. (2004) Force measurements in E-cadherin-mediated cell doublets reveal 3. Solnica-Krezel L, Sepich DS (2012) Gastrulation: Making and shaping germ layers. rapid adhesion strengthened by actin cytoskeleton remodeling through Rac and Annu Rev Cell Dev Biol 28:687–717. Cdc42. J Cell Biol 167:1183–1194. 4. Voiculescu O, Bodenstein L, Lau I-J, Stern CD (2014) Local cell interactions and self- 28. Lele Z, et al. (2002) parachute/n-cadherin is required for morphogenesis and main- amplifying individual cell ingression drive amniote gastrulation. Elife 3:e01817. tained integrity of the zebrafish neural tube. Development 129:3281–3294. 5. Carmany-Rampey A, Schier AF (2001) Single-cell internalization during zebrafish 29. Tan RZ, Chiam KH (2014) Computational modeling reveals that a combination of gastrulation. Curr Biol 11:1261–1265. chemotaxis and differential adhesion leads to robust cell sorting during tissue pat- PLoS One 6. David NB, Rosa FM (2001) Cell autonomous commitment to an endodermal fate and terning. 9:e109286. behaviour by activation of Nodal signalling. Development 128:3937–3947. 30. Migeotte I, Grego-Bessa J, Anderson KV (2011) Rac1 mediates morphogenetic re- Development 7. Kane D, Adams R (2002) Life at the edge: Epiboly and involution in the zebrafish. sponses to intercellular signals in the gastrulating mouse embryo. 138: – Results Probl Cell Differ 40:117–135. 3011 3020. 8. Townes PPL, Holtfreter J (1955) Directed movements and selective adhesion of em- 31. Montero J-A, et al. (2005) Shield formation at the onset of zebrafish gastrulation. Development – bryonic amphibian cells. J Exp Zool 128:53–120. 132:1187 1198. 9. Steinberg MS (2007) Differential adhesion in morphogenesis: A modern view. Curr 32. Schäfer G, Narasimha M, Vogelsang E, Leptin M (2014) Cadherin switching during the formation and differentiation of the Drosophila mesoderm–Implications for Opin Genet Dev 17:281–286. epithelial-to-mesenchymal transitions. J Cell Sci 127:1511–1522. 10. Takeichi M (1995) Morphogenetic roles of classic cadherins. Curr Opin Cell Biol 7: 33. Hardy KM, Yatskievych TA, Konieczka J, Bobbs AS, Antin PB (2011) FGF signalling 619–627. through RAS/MAPK and PI3K pathways regulates cell movement and gene expression 11. Krieg M, et al. (2008) Tensile forces govern germ-layer organization in zebrafish. Nat in the chicken primitive streak without affecting E-cadherin expression. BMC Dev Biol Cell Biol 10:429–436. 11:20. 12. Ninomiya H, et al. (2012) Cadherin-dependent differential cell adhesion in Xenopus 34. Nieman MT, Prudoff RS, Johnson KR, Wheelock MJ (1999) N-cadherin promotes causes cell sorting in vitro but not in the embryo. J Cell Sci 125:1877–1883. motility in human breast cancer cells regardless of their E-cadherin expression. J Cell 13. Row RH, et al. (2011) Completion of the epithelial to mesenchymal transition in ze- Biol 147:631–644. brafish mesoderm requires Spadetail. Dev Biol 354:102–110. 35. Nguyen T, Mège RM (2016) N-cadherin and fibroblast growth factor receptors crosstalk 14. Viotti M, Nowotschin S, Hadjantonakis A-K (2014) SOX17 links gut endoderm mor- in the control of developmental and cancer cell migrations. Eur J Cell Biol 95:415–426. phogenesis and germ layer segregation. Nat Cell Biol 16:1146–1156. 36. Gritsman K, et al. (1999) The EGF-CFC protein one-eyed pinhead is essential for nodal 15. Warga RM, Nüsslein-Volhard C (1999) Origin and development of the zebrafish en- signaling. Cell 97:121–132. doderm. Development 126:827–838. 37. Ho RK, Kimmel CB (1993) Commitment of cell fate in the early zebrafish embryo. 16. Keller PJ, Schmidt AD, Wittbrodt J, Stelzer EHK (2008) Reconstruction of zebrafish early Science 261:109–111. Science – embryonic development by scanned light sheet microscopy. 322:1065 1069. 38. Westerfield M (2007) The Zebrafish Book. A Guide for the Laboratory Use of Zebrafish Annu Rev Cell Dev Biol 17. Jaffe AB, Hall A (2005) Rho GTPases: Biochemistry and biology. (Danio rerio). (Univ of Oregon Press, Eugene, OR), 5th Ed. – 21:247 269. 39. Peyriéras N, Strähle U, Rosa F (1998) Conversion of zebrafish blastomeres to an en- 18. Da Costa SR, et al. (2003) Impairing actin filament or syndapin functions promotes dodermal fate by TGF-beta-related signaling. Curr Biol 8:783–786. accumulation of clathrin-coated vesicles at the apical plasma membrane of acinar 40. Tahinci E, Symes K (2003) Distinct functions of Rho and Rac are required for con- Mol Biol Cell – epithelial cells. 14:4397 4413. vergent extension during Xenopus gastrulation. Dev Biol 259:318–335. 19. Chen S, Kimelman D (2000) The role of the yolk syncytial layer in germ layer pat- 41. Djiane A, Riou J, Umbhauer M, Boucaut J, Shi D (2000) Role of frizzled 7 in the reg- terning in zebrafish. Development 127:4681–4689. ulation of convergent extension movements during gastrulation in Xenopus laevis. 20. Hong S-K, Dawid IB (2008) Alpha2 macroglobulin-like is essential for liver develop- Development 127:3091–3100. ment in zebrafish. PLoS One 3:e3736. 42. Babb SG, Marrs JA (2004) E-cadherin regulates cell movements and tissue formation in 21. Thisse B, et al. (2001) Expression of the zebrafish genome during embryogenesis. ZFIN early zebrafish embryos. Dev Dyn 230:263–277. direct data submission. 43. LaMora A, Voigt MM (2009) Cranial sensory ganglia neurons require intrinsic 22. Hatta K, Takeichi M (1986) Expression of N-cadherin adhesion molecules associated N-cadherin function for guidance of afferent fibers to their final targets. with early morphogenetic events in chick development. Nature 320:447–449. Neuroscience 159:1175–1184. 23. Oda H, Tsukita S, Takeichi M (1998) Dynamic behavior of the cadherin-based cell-cell 44. Bajoghli B, Aghaallaei N, Heimbucher T, Czerny T (2004) An artificial promoter construct adhesion system during Drosophila gastrulation. Dev Biol 203:435–450. for heat-inducible misexpression during fish embryogenesis. Dev Biol 271:416–430. 24. Wheelock MJ, Shintani Y, Maeda M, Fukumoto Y, Johnson KR (2008) Cadherin 45. Aoki TO, et al. (2002) Molecular integration of casanova in the Nodal signalling switching. J Cell Sci 121:727–735. pathway controlling endoderm formation. Development 129:275–286.

10148 | www.pnas.org/cgi/doi/10.1073/pnas.1708116114 Giger and David Downloaded by guest on September 29, 2021