Construction of a Vertebrate Embryo from Two Opposing Morphogen Gradients

See allHide authors and affiliations

Science  04 Apr 2014:
Vol. 344, Issue 6179, pp. 87-89
DOI: 10.1126/science.1248252

Designer Embryo

Numerous signaling pathways have been implicated in controlling early vertebrate embryogenesis. P.-F. Xu et al. (p. 87) identify the minimal set of factors necessary to get uncommitted cells to organize a complete embryo. Two opposing gradients of the growth factors Nodal and Bone Morphogenetic Protein were sufficient to instruct zebrafish embryonic pluripotent cells to organize a complete embryo, not only in vivo but also in vitro. These findings may provide guidance for regenerative medicine studies aimed at constructing tissues and organs in vitro from cultured pluripotent cells.


Development of vertebrate embryos involves tightly regulated molecular and cellular processes that progressively instruct proliferating embryonic cells about their identity and behavior. Whereas numerous gene activities have been found to be essential during early embryogenesis, little is known about the minimal conditions and factors that would be sufficient to instruct pluripotent cells to organize the embryo. Here, we show that opposing gradients of bone morphogenetic protein (BMP) and Nodal, two transforming growth factor family members that act as morphogens, are sufficient to induce molecular and cellular mechanisms required to organize, in vivo or in vitro, uncommitted cells of the zebrafish blastula animal pole into a well-developed embryo.

The formation of vertebrate embryos depends on the activity of an organizing center, corresponding to the dorsal blastopore lip in amphibians and known as the Spemann-Mangold or dorsal organizer (1). When grafted ventrally into a host, this organizer results in the formation, at the site of the graft, of a secondary embryonic axis. The molecular nature of the activity carried by this dorsal tissue has been identified (2), and its main components are dorsally secreted factors that act as antagonists of ventral morphogens and help to establish a ventral-to-dorsal gradient of their activity (fig. S1A). However, when placed in a neutral environment such as the blastula animal pole, the Spemann organizer has very limited organizing activity, leading only to induction of axial mesendodermal tissues (3).

We previously established that, in zebrafish, the organizing activities controlling the development of the embryo are not restricted to the dorsal organizer but are distributed all along the embryonic margin (3, 4) and result from the combined activity of two signaling pathways: bone morphogenetic protein (BMP) and Nodal. We found that the organizing activity of each portion of the embryonic margin varies depending on the BMP/Nodal ratio of activities (fig. S1B): A high BMP/Nodal ratio organizes the tail; a low BMP/Nodal ratio organizes the posterior head, whereas intermediate ratios of activity induce formation of the trunk (4). Based on these observations, we hypothesized that exposing uncommitted embryonic cells to a continuous variation of BMP/Nodal ratios may be sufficient to organize a complete embryonic axis (fig. S1C).

To test this hypothesis, we engineered opposing gradients of BMP and Nodal by injecting their corresponding mRNAs, at the 128-cell stage, into two different animal pole blastomeres that give rise to distinct clones of cells secreting these factors (Fig. 1A). The animal pole of the blastula was chosen as the territory to instruct because it contains all elements necessary to mediate the molecular and cellular responses to BMP and Nodal stimulation (3, 4). Additionally, because the animal pole is distant from the margin, the influence of the primary axis and of maternal determinants is expected to be minimal.

Fig. 1 BMP and Nodal induce a secondary embryonic axis at the animal pole.

(A) Injection of Nodal and BMP mRNAs in two different animal pole blastomeres at the 128-cell (128c) stage results in (B) formation of a secondary embryonic axis (ii) expressing sonic hedgehog a from the midbrain to the tip of the tail. (C to E) Secondary embryonic axes including anterior head and displaying antero-posterior (a-p) orientation (red arrow) (C) parallel, (D) perpendicular, and (E) antiparallel to a-p orientation (blue arrow) of the primary axis (i). e, eye; fp, floorplate; h, heart; hg, hatching gland; n, notochord; mb, midbrain. Images in (B) to (E) are composite photos from a stack of images made at different planes of focus. hpf, hours postfertilization.

In support of our hypothesis, these two secreting centers organize the animal pole cells (see supplementary materials and methods) into a secondary embryonic axis that forms at the animal pole (Fig. 1, B to E), solely from animal pole cells (fig. S2). These ectopic axes contain tissues and organs present in the primary axis and extend from the forebrain to the tip of the tail (Fig. 1B). In most cases, the primary and secondary axes fuse in the cephalic region, where animal pole cells are recruited by both growing structures. However, in some cases (1.3%; n = 1012 embryos), clones are in such a position that the two embryonic axes do not fuse (Fig. 1, C to E), with each displaying a forebrain, eyes, and a beating heart and exhibiting spontaneous myotomal contractions indicative of a functional nervous system (movie S1). Although the antero-posterior (A-P) axis of the primary embryo always parallels the animal-vegetal (An-Vg) axis of the egg, we found no correlation between the An-Vg axis of the egg and the A-P axis of the secondary embryo, which can be parallel, perpendicular, or even antiparallel to the A-P axis of the primary embryo and, therefore, to the An-Vg axis of the egg (Fig. 1, C to E). This demonstrates that, in zebrafish, there is no intrinsic information present in the egg or in the early embryo determining the orientation of the A-P axis that cannot be reversed by application of appropriate signals.

To understand the relative contribution of the two signaling pathways to the organization of the secondary embryonic axis, we examined the consequence of adding BMP or Nodal individually. At the blastula stage, BMP signaling is already active at the animal pole (5); adding more BMP to this region has very little effect (fig. S3). Conversely, at the blastula stage, stimulating the animal pole with Nodal results in a thickening of the animal pole blastoderm (Fig. 2A). At the onset of gastrulation, the central part of the resulting protrusion internalizes (Fig. 2, B and C), forming a blastopore with a circular, radially symmetrical blastopore lip where mesodermal cells involute (movie S2).

Fig. 2 Nodal induces formation of a blastopore where ectopic gastrulation occurs.

(A) Thickening of the blastoderm (arrowheads) induced by ectopic Nodal signaling. (B) Formation of a blastopore (bl) at gastrulation. (C) Internalization of the Nodal-secreting cells [Nodal plus red fluorescent protein (RFP)]. (D) Blastopore induced by Nodal after STAT3 depletion by morpholino (STAT3 Mo) injection. (E and F) Nodal induces expression of liv1 in the wild type (E), but not in STAT3-depleted embryos (F). (G and H) Expression of genes induced by Nodal at the animal pole revealed by double-color in situ hybridization. Arrows in (H) indicate the spreading of prechordal plate cells expressing frzb. The developmental stage is indicated in the lower left corner of each panel. ep, epiboly. Probes are indicated in the lower right corner in (E) to (H). (A to D and G) Lateral view; (E, F, H) animal pole view. Scale bar in (A) corresponds to 100 μm for (A), (B), (E), and (F) and 40 μm for (C) and (D).

At blastula and gastrula stages, epiboly disperses cells of the animal pole over the ectoderm. However, in the presence of a Nodal-expressing clone, the surrounding animal pole cells do not spread but converge toward the center of the Nodal-secreting source (fig. S4). It is known that the guidance cues for convergence movements in the zebrafish gastrula are controlled by janus kinase/signal transducer and activator of transcription 3 (JAK/STAT3) signaling (6, 7). Activation of STAT3 is independent of Nodal signaling but dependent on the maternal β-catenin signaling pathway (6). Because ectopic Nodal signaling induces an attracting center at the animal pole, we hypothesized that STAT3 may be activated in that domain. Consistent with this, in STAT3 morphant embryos, ectopic Nodal signaling fails to induce radial convergence movements, whereas internalization of the mesendoderm is not affected (Fig. 2D). Furthermore, Nodal induction of expression of liv1 (Fig. 2E), a downstream target of STAT3 (8), is abolished in STAT3 morphant embryos (Fig. 2F), demonstrating that although Nodal is not required for activation of STAT3 in the dorsal domain of the blastula (7), it is sufficient to induce this activation at the animal pole.

In zebrafish, Nodal is known to be essential for inducing the organizing activity carried by the dorsal gastrula margin (911). Accordingly, stimulation of animal pole cells by Nodal induces expression of dorsal and dorso-lateral marginal genes (fig. S5) but never induces markers of lateral or ventral cell identities such as eve1 (Fig. 2G). At the onset of gastrulation, the expression of genes induced at the animal pole by Nodal is organized in concentric circles (fig. S5), reflecting the morphogenic activity of Nodal (12). At late gastrula stage, a notochordal domain expressing ntl extends out of the animal pole as a result of radial convergence and extension movements (Fig. 2G), and the involuted prechordal plate cells expressing frzb have reached the yolk syncytial layer and spread in all directions in a radially symmetrical manner (Fig. 2H). Thus, engineering a Nodal activity gradient at the animal pole results in the formation of a radially symmetric structure possessing dorsal identity.

Providing a gradient of BMP opposed to the Nodal activity gradient (Fig. 3A) converts the radially symmetrical structure induced by Nodal alone into a bilaterally symmetric embryonic axis. Time-lapse analysis reveals that, at the blastula stage (movie S3), the protrusion induced in the presence of both BMP and Nodal gradients is morphologically identical to the protrusion induced by Nodal alone (Figs. 2A and 3A). However, at gastrulation, we observe a thinning of the blastopore lip close to the BMP-secreting clone and a simultaneous thickening of the opposite side (Fig. 3B), suggesting that cells are migrating away from the BMP source. This is likely to reflect the repulsive effect of BMP on migrating lateral mesodermal cells (13). In addition, and in accordance with its function in dorso-ventral (D-V) axis formation (1416), BMP patterns the blastopore lip induced by Nodal, defining a ventral domain close to the BMP-secreting center and progressively more dorsal identities with increasing distance from the BMP source. Consequently, at the gastrula stage, dorsal marginal markers are restricted to the domain of the blastopore lip furthest from the BMP-secreting clone (Fig. 3, C to E), whereas ventral- and lateral-specific genes are induced close to the BMP-secreting center (Fig. 3F and fig. S6).

Fig. 3 BMP breaks the radial symmetry and patterns the blastopore lip induced by Nodal.

(A) BMP- and Nodal-secreting clones induce (B) a blastopore with an asymmetrical blastopore lip. (C to F) Expression of genes induced at the animal pole by Nodal and BMP. The developmental stage is indicated in the lower left corner of each panel; probes are indicated in the lower right corner in (C) to (F). Brown cells in (C) to (F) are cells secreting BMP. (A and B) Lateral view; (C to F) animal pole view. Scale bar in (A) corresponds to 100 μm for (A) and 40 μm for (B) to (F).

Whereas D-V patterning requires BMP activity (16), establishment of the A-P polarity has been shown to depend on Nodal (17). Accordingly, the A-P polarity of the secondary axes induced by Nodal or by Nodal and BMP is in the opposite orientation to that of the primary embryonic axis (fig. S7, A and B). This is probably the consequence of the expression, in the induced ectopic blastopore lip, of FGF8a (18) and Wnt8a (19), which act as posteriorizing morphogens (20, 21) (fig. S7, C to F). Although these posteriorizing factors control the initial A-P polarity, the direction of extension of the induced secondary axis depends only on the position of the BMP-secreting clone relative to the blastopore lip induced by Nodal and is completely independent of the orientation of the primary embryonic axis (fig. S7, G and H, and fig. S8).

All of these data support the conclusion that formation of the secondary embryonic axis that we generate occurs without any contribution from signals that organize and pattern the primary axis and that it depends only on the clones secreting BMP and Nodal that we engineered. To further demonstrate this, we tested the ability of BMP and Nodal to induce embryonic development from blastula animal pole explants cultured in vitro and, therefore, independent of the primary axis.

Explantation of animal pole cells is performed at the 512-cell stage [midblastula transition (MBT)] (Fig. 4A). At that stage, animal pole cells of the zebrafish embryo are all pluripotent, undifferentiated, and equivalent (22). Notably, nuclei of animal pole cells are completely devoid of β-catenin (23), which accumulates in the nuclei of only a small number of dorso-marginal cells (24, 25). This is different from what is observed for amphibian embryos that display a preexistent pattern at the animal pole of the blastula (26) that depends on the endogenous D-V gradient of β-catenin (24).

Fig. 4 In vitro construction of embryonic axes with opposing gradients of BMP and Nodal.

(A) Animal pole injection and explantation. c, cells. (B) Uninjected explants. (C to E) Explants from embryos injected with both BMP and Nodal mRNAs analyzed at the gastrula stage (C) for the morphology (left) and by in situ hybridization with endoderm (sox32), prechordal (frzb), notochord (shha), and epidermal (foxi1) markers or observed at 24 hpf (D and E) and analyzed by in situ hybridization with retina (ret, vsx2), spinal cord neurons (ne, vsx2), forebrain (fb, emx1), hindbrain (hb, egr2b), and somite (so, myoD) markers. bl, blastopore; n, notochord; nt, neural tube. Scale bar in (A) corresponds to 200 μm for (A), 80 μm for (B) and (C), and 100 μm for (D) and (E).

The animal pole of uninjected embryos, explanted at the MBT and placed in culture medium, does not differentiate any embryonic tissues or organs (Fig. 4B). Explants become spherical and, after ~1 hour, differentiate a morphologically visible enveloping layer (EVL). Three hours later, an internal cavity appears within the explant, reminiscent of a blastocoel, a fluid-filled central cavity of the blastula, which is observed in most vertebrate species but is absent in zebrafish (27).

Our analysis of genes expressed in these explants at time points corresponding to gastrulation in control embryos (fig. S9) reveals that animal pole cells at or before the time of explantation have received no signals induced by maternal dorsal determinants. After 1 day of culture, no morphologically visible structures are present, and we only observed a rupture of the EVL with extrusion of some of the enclosed embryonic cells (Fig. 4B). Stimulation by injection of Nodal mRNA into one animal pole blastomere before explantation later induces gastrulation with formation of a radially symmetrical blastopore lip (movie S4). When secreting centers for both Nodal and BMP are engineered, gastrulating explants display an asymmetrical blastopore lip, contain cells from the three germ layers, and are patterned along clear A-P and D-V axes (Fig. 4C). After 1 day of culture, these explants differentiate into embryoids that display morphologically recognizable structures such as the forebrain, neural tube, notochord, and somites (Fig. 4D). These embryoids express organ-specific molecular markers (Fig. 4E); their neural tubes are patterned along the A-P axis, and they often display bilateral symmetry, as illustrated by vsx2 expression, which reveals the presence of two eyes (Fig. 4E).

Similar observations have been made after the fusion of two animal pole explants: one uninjected explant and the other explant injected with BMP and Nodal mRNAs (fig. S10). In this condition, all mesendodermal tissues derive from the injected explant, whereas cells of the uninjected explant differentiate anterior ectodermal derivatives. The ability of the BMP and Nodal signals to organize the fused explants (that initially contain two animal poles with An-Vg axes in opposite orientation) into a single embryoid with a clear A-P axis provides proof that a strongly biased preexistent pattern is not present within the animal pole region of the zebrafish blastula.

Our study demonstrates that stimulation of uncommitted cells at the blastula animal pole with opposing gradients of BMP and Nodal is sufficient to initiate the principal molecular and cellular processes necessary to organize a complete embryonic axis. All other signaling pathways required to achieve full embryonic development are induced and regulated in response to the two initial, experimentally engineered signals. Therefore, our findings establish a baseline for the minimal signaling requirements for early embryonic development and provide a framework for future studies in the field of regenerative medicine that are aimed at constructing tissues and organs in vitro from populations of cultured pluripotent cells.

Supplementary Materials

Material and Methods

Figs. S1 to S10

References (2832)

Movies S1 to S4

References and Notes

  1. Acknowledgments: We thank L. Solnica-Krezel for fruitful discussion; R. Bloodgood, R. Keller, and A. Sutherland for careful reading of the manuscript; S. Snyder for technical assistance; and S. Vecchio for taking care of the fish. K.F.F.-L. was supported by the European Molecular Biology Organization. This work was supported by funds from the University of Virginia.
View Abstract

Navigate This Article