Who moves whom during primitive streak formation in the chick embryo

Abstract

Gastrulation is a critical stage in the development of all vertebrates. During gastrulation mesendoderm cells move inside the embryo to form the gut, muscles, and skeleton. In amniotes the mesendoderm cells move inside the embryo through a structure known as the primitive streak, extending from the posterior pole anterior through the midline of the embryo. Primitive streak formation involves large scale cell flows of a layer of highly polarized epithelial epiblast cells. The epiblast is separated from a lower layer of hypoblast cells through a well developed basal lamina. Recent experiments in which in vivo extracellular matrix dynamics was followed via labeling with fibronectin specific fluorescent antibodies and time-lapse microscopy have suggested that extracellular matrix dynamics essentially coincides with the observed epiblast cell displacements (Zamir et al., 2008, PLoS Biol 6, e247). These observations raise the important question of who moves whom and where do cells derive traction. We discuss these matters and their implications for our understanding of the mechanisms underlying cell flows during primitive streak formation in the chick embryo.

One of the big open problems in the study of early development of higher animals is to understand the cellular mechanisms underlying the large scale cell displacements that take place during gastrulation and the signals that control them. The process of gastrulation is studied in many invertebrate and vertebrate species and in amniotes, especially in chick and mouse (Stern, 2004). Chick embryos are flat and can be successfully cultured in vitro and are amenable to extensive experimental manipulation. At the time of egg laying the embryo contains around 40.00 cells that are arranged in an almost radially symmetric pattern. The embryo consists of an upper layer of polarized epithelial cells, the epiblast, which at the periphery is continuous with a several cell layer thick collection of large yolk rich, mesenchymal cells, an area known as the area opaca, which will develop into extra-embryonic structures. The central inner essentially on cell layer thick area is known as the area pellucid and the cells in this region will give rise to the embryo. The area opaca and area pellucida are separated by a group of distinctively shaped epithelial cells, the marginal zone. During early development a sickle shaped assembly of loosely associated cells underneath the epiblast can be distinguished at posterior pole of the area pellucida, a structure known as Koller’s Sickle. Shortly afterward the mesendoderm starts to form by differentiation of cells in the epiblast overlying Koller’s Sickle, in response to signals coming from the surrounding marginal zone and area opaca. Gastrulation starts when cells that will form the future mesendoderm move into the midline of the embryo to form the primitive streak. The primitive streak is macroscopically visible as a darker area formed by local stacking of epiblast cells on top of each other. Streak formation starts at the posterior pole of the epiblast followed by an elongation in anterior direction. When the streak has extended about halfway over the epiblast, the deeper cells of the streak start to move outward in between the epiblast and the hypoblast to form the gut, muscles, and skeleton.

At the time of egg laying, the epiblast cells already are strongly polarized. They form apically localized adherens and tight junctions and express several basal membrane components such as fibronectin, laminin, and most likely several integrins in their basal membranes (Sanders, 1982; Zagris, 2001). Attached to this sheet of epithelial cells are small groups of rounded cells that form the primary hypoblast; these hypoblast cells derive from the epiblast via an ingression process known as polyingression (Eyal-Giladi and Kochav, 1976; Weinberger and Brick, 1982a). The movements associated with gastrulation begin 4 to 5 h after incubation of fertilized eggs at 37 °C. The first observable movements are associated with the formation of the secondary hypoblast, which forms in a posterior to anterior direction [Fig. 1A]. Some of the cells that form the secondary hypoblast appear to derive from cells of Koller’s Sickle, which are moving forward (Eyal-Giladi et al., 1992; Spratt and Haas, 1960). During hypoblast development these cells flatten and fuse to form an epithelial sheet of cells, which during its forward extension also incorporates cells from the primary hypoblast (Low and McClugage, 1993). During the early stages of development there is considerable cell division both in the epiblast and in the hypoblast certainly contributing to the growth and expansion of the embryo [Fig 1B]. Evidence has, however, been presented claiming that the hypoblast can form in the absence of significant cell division, suggesting that cell movement and possibly ingression from the epiblast may be sufficient to account for hypoblast formation (Weinberger and Brick, 1982b).

Figure 1. Structure of early chick embryo and movement patterns of cells in epiblast.

(A) Overview image of a 3 h incubated (EG XII) embryo showing the hypoblast cells stained in green (as detected with a HNK1 antibody, which recognizes a cell surface carbohydrate epitope strongly expressed on hypoblast cells). The anterior posterior axis of the embryo is oriented vertically with posterior pole of the embryo located in the bottom of the image. (B) High magnification image of the epiblast showing the cell boundaries of the epiblast cells in green [phalloidin staining of filamentous actin, mainly associated with the apical adherens junctions, see Fig. 2C] and the nuclei of these cells in red (Propidium-iodide staining of DNA). Note the presence of four mitotic cells indicated by white arrows. Cells in mitosis round off, retract, in between other cells in the epiblast and in doing so pull the apical sides of adjacent cells into typical rosette structures, where more than three neighboring cells meet. (C) Section through the embryo shown in (B), showing individual epiblast cells transfected with GFP. Note the tall shape of the epiblast cells connecting overlaying a discontinuous layer of hypoblast cells. (D) Cell flow patterns observed by tracing individual GFP transfected cells in the epiblast over the course of 10 h. Movement direction is from yellow to green. The green part of the track is the movement recorded over the last hour of the experiment.

STREAK FORMATION IS CHARACTERIZED BY LARGE SCALE CELL FLOWS

During the expansion of the hypoblast, cells in the overlying epiblast start to move in typical flowlike patterns; epiblast cells at the posterior half along the boundary of the area opaca and area pellucid start to move toward the posterior pole where the flows from both directions merge and start to move through the middle of the embryo to anterior along the central midline to form the streak [Fig. 1D]. These movements were originally followed by tracing of carbon particles attached to the cells and have been named polonaise movements (Graeper, 1929; Spratt and Haas, 1960). Once the streak has extended halfway over the epiblast and the cells in the epiblast are starting to ingress, cells in the epiblast now also start to move from lateral positions toward the streak and once they reach the streak these epithelial cells undergo an epithelial to mesenchymal transition (EMT) and also start to ingress. Mesoderm and endoderm cells move into the embryo and migrate as individual cells over large distances in between the epiblast and the hypoblast to form various mesodermal and endodermal structures (Keller, 2005). Once the cells have undergone EMT they move by getting traction from each other and the extracellular matrix (Yang et al., 2008). The movement of individual mesenchymal cells is in principle better understood, since it more close resembles movement of cells on a substrate, which is being studied in tissue culture in many laboratories. More recently the movement of the cells in the epiblast has been studied in more detail by DiI labeling of small groups of cells and by transfection of randomly scattered cells in the epiblast [Fig. 1D]. These new techniques, especially the ability to transfect cells with GFP, is starting to allow the study of the detailed cell behaviors associated with streak formation (Chuai et al., 2006; Cui et al., 2005; Voiculescu et al., 2007). The mechanisms underlying the movement of the epithelial cells resulting in the early phases of streak formation are still not very well understood and there are still many unresolved questions:

• How do epithelial cells of the epiblast move in embryos to form the streak?

• How much of the observed movement is active and how much is passive?

• What are the signals that coordinate the behavior of the cells?

• What are the relationships between signaling and movement?

The large scale cell flows that have been described for both the hypoblast and epiblast are relative to the outside world. There is little doubt that some of the observed flows must have their basis in active cell motility, but at the outset it is not clear whether all the observed flows are the result of active individual cell movements or whether some cell movements are passive displacements, possibly largely the result of changes in tissue geometry elsewhere in the embryo. Until now three distinct mechanisms had been proposed to drive streak formation (Chuai and Weijer, 2007; Wei and Mikawa, 2000):

• Localized oriented cell divisions (Wei and Mikawa, 2000),

• cell-cell intercalation in the posterior streak resulting in localized convergent extension (Lawson and Schoenwolf, 2001; Voiculescu et al., 2007), and

• chemotactic cell movement in response to a combination of attractants and repellents (Chuai and Weijer, 2007).

So far experimental and theoretical evidence has been presented that streak formation does not require cell division at least to proceed through the early stages of streak development (Bodenstein and Stern, 2005; Chuai et al., 2006). Both other mechanisms, cell-cell intercalation and chemotaxis, require active cell movements, either small active local cell movements that add up to result in long range tissue deformations in the case of a cell-cell intercalation mechanism or large scale active cell movements in the case of a chemotactic mechanism (Fig. 2). Also, the mechanisms by which the cells move are not yet clear. Cells can move either by getting traction from neighboring cells, through pulling on neighboring cells, or they can move by extension and retraction of lamellipodia at their basal surface where they get traction from the basal lamina (Keller, 2005). Experiments in mouse embryos and embryoid bodies have shown that especially laminins play important roles in early embryonic cell polarization and tissue organization at preimplantation and early primitive streak stages and that laminins are required for formation of fibronectin and collagen IV fibrils that constitute important structural components of the basal lamina (reviewed in Li et al., 2003; Miner et al., 2004; Miner and Yurchenco, 2004).

Figure 2. Possible modes of cell behavior explaining the observed cell flow patterns.

(A) Flow patterns observed in a liquid contained in a circular domain in response to a locally applied force (the red arrow). Although the force is applied locally, the fluid moves everywhere as result of viscous interactions. In biological tissues these long range interactions could be mediated by adhesive interactions between cells or interactions of the cells with the matrix. (B) Schematic cross section through an early embryo. Epiblast and hypoblast cells are in contact with a complex basal lamina through specific lamina integrin interactions (green double arrows). In principle basally extended lamellipodia can get traction from this matrix, through coupling to integrins to the internal actin cytoskeleton, as long as the matrix is stationary with respect to the cells. If the cells cannot get traction from the matrix, they can only get traction from each other either through protrusions that make cell-cell contacts or through directional relaxation and contraction of the actin cables associated with the adhesive junctional complexes (red double arrows). (C) Scheme outlining how retraction of specific junctions (red) in one direction (blue arrows) followed by relaxation of these junctions is a direction perpendicular to the original direction of contraction (blue arrows) results in a tissue deformation and reorientation. This process involves rosette formation as an intermediate stage.

THE ROLE OF MATRIX DYNAMICS IN EPIBLAST CELL DISPLACEMENTS

Recently Little and co-workers have started to investigate the dynamics of the extracellular matrix during the early phases of gastrulation and their experimental results surprisingly suggest that there is little epiblast cell movement relative to the matrix (Czirok et al., 2004, 2006; Zamir et al., 2008). The observations that the matrix moves in synchrony with the observed cell displacements are based on the detection the dynamics of fibronectin detected by injection of fluorescently labeled antibodies against this molecule (Zamir et al., 2008). These experiments show that the fluorescent label moves with similar dynamics as the epiblast cells, which are detected by expression of a GFP based nuclear marker. At first glance these results would suggest that either the cells take their matrix with them, or the matrix moves and carries the cells on top with it. In any case these results would suggest that it would be difficult for the cells to get traction from the matrix to move. Furthermore, these findings that the matrix moves with the same speed as the cells make it difficult to envisage that the matrix could be a carrier of graded chemical signals that could instruct the cells on where to move. Therefore, these observations could have far reaching consequences for our understanding of gastrulation movements and therefore these results and their implications require a further analysis.

There is abundant evidence that cells can modulate the extracellular matrix at least on a small scale by localized remodeling, for instance to accommodate growth of the embryo. This inevitably will involve making and breaking interactions between various matrix components. It will, however, be important to unequivocally demonstrate whether the extracellular matrix moves on a large scale and that the antibody label technique does accurately reflect the real dynamics of the matrix. Fibronectin after synthesis and secretion is known to exist in several distinct states, a soluble state, a cell bound state, and a state of incorporation into large fibrilar complexes (Mao and Schwarzbauer, 2005; Schwarzbauer and Sechler, 1999). Epiblast and hypoblast cells secrete soluble fibronectin, which is thought be in a globular form. It is believed that secreted fibronectin binds to α₅β₁ integrins, specific transmembrane cell surface receptors with an extracellular fibronectin binding domain. Fibronectin binding results in integrin activation, clustering, and activation of the actin cytoskeleton, which somehow unfolds the globular fibronectin molecule, allowing it to bind to and interact with other fibronectin molecules to form fibrils, which in turn interact with many other extracellular matrix components (Daley et al., 2008; Mao and Schwarzbauer, 2005). The epitope binding sites of the antibodies used in this study have not been mapped, but the antibodies used were divalent and therefore have the ability to cross link fibronectin molecules, possibly leading to a failure of suitable processing and assembly of fibronectin into functional fibrils. It seems possible that the labeling techniques used somehow immobilized fibronectin to the cell surface, resulting in cells apparently taking some of the matrix with them. This caveat could be addressed by repeating the experiments with monovalent Fab antibody fragments, which can still bind fibronectin but have lost their ability to cross link fibronectin molecules. Alternatively it would be possible to inject fluorescently labeled fibronectin, which will then become incorporated into the fibrillar network by the action of the cells as has been shown in several systems most prominently in extracellular matrix remodeling during lung morphogenesis (Larsen et al., 2006). Another way to answer this question would be to express GFP tagged fibronectin in epiblast cells and determine whether the fibronectin synthesized by cells stays with the cells or whether it is left behind in a trail as the cells move (Ohashi et al., 2002). Furthermore, it would be desirable to see whether other important protein components of the basal lamina such as laminin and especially collagen IV show the similar behavior as fibronectin, i.e., move along with the cells.

In case further experiments confirm that the matrix indeed undergoes large scale movements with similar dynamics as the cells, then there will be many questions that need to be resolved. Major questions are do the cells move actively and take the matrix with them, or is the matrix being deformed, possibly by cells elsewhere in the embryo, and does the matrix carry the epiblast cells with it, i.e., is epiblast cell displacement passive? It could be that the cells take their own matrix with them and that the main role of the matrix is signaling, for instance to stabilize apical basal polarity or at best to locally coordinate cell behaviors, but is not required for individual cells to obtain traction for active movement. If the cells move active and for some as yet unknown reason take their matrix with them, then cells would have to gain traction from other cells through cell-cell contacts. Cells could rearrange relative to other cells through active modulation of the actin filamentous network associated with adherens junctions [Fig. 2B]. Studies in Drosophila have suggested that the cell movements associated with germband extension are based on a cell-cell intercalation mechanism that depends on local contraction of the cell-cell junction associated actin network in a myosin II dependent process (Bertet et al., 2004; Blankenship et al., 2006). This contraction is then followed by relaxation of the junctions in a direction perpendicular to the direction of the original contraction. If these contractions occur in a coordinated manner in neighboring cells, then this can result in local or large scale tissue deformations [Fig. 2C]. Alternatively, the cells could polarize and use cellular protrusions to get traction from each other as has been suggested to occur during intercalation of mesoderm cells during convergent extension in Xenopus (Keller, 2005). Studies in Xenopus have suggested that activation of α₅β₁ integrins signaling through binding of the fibronectin of the basal lamina may suppress protrusive activity at the level of the basal lamina and thereby prevent cells moving on top of each other, thus maintaining an epithelial structure. Inhibition of integrin function through application of function blocking antibodies that prevent the binding to fibronectin result in excessive protrusive activity and loss of stability of the epithelial sheet (Davidson et al., 2008, 2006). In Xenopus it has recently been shown that inhibition of fibronectin fibril assembly by overexpression of an N terminal 70 K fibronectin domain involved in fibril formation results in randomization of the cell division planes and thereby results in a thickening of the animal cap, which prevents the radial intercalation that occurs normally, which results in a failure of epiboly and therefore proper closure of the blastoporal lip (Rozario et al., 2008). These experiments suggest that ligand dependent integrin activation is an important signaling event controlling protrusive activity and orientation of the cell division planes. Thus contact with the basal lamina is thought to be important maintaining the polarized state of the epithelium, and, in the case of the chick epiblast could play an important role in stabilizing the epithelial structure of the epiblast, functions which would be difficult to reconcile with the idea that the cells take their matrix with them. In the chick embryo both laminin and fibronectin are expressed by epiblast cells in the freshly laid egg. It has been described that blocking laminin with function blocking antibodies results in changed movements and a failure of normal streak formation (Zagris and Chung, 1990). Injection of fibronectin function blocking antibodies has been described to result in the ingression of epiblast cells into the streak, but a failure of mesenchymal cells to move away from the streak, suggesting that these cells need fibronectin to get traction to move (Harrisson et al., 1993). To assess the role of fibronectin in gastrulation, detailed loss of function experiments, i.e., deletion of fibronectin, and other extracellular matrix components are needed to assess the role of the cell matrix interactions in cell movement during streak formation.

Finally, if the matrix moves with the same speed as the cells, it is possible in theory that the matrix acts as a type of conveyor belt moving the cells. This would imply that either the matrix has to be generated by the cells somewhere in the embryo and pulled in and degraded somewhere else. Since the cells not only move apart but actually move in two counter-rotating cell flows, it is difficult to see how this possibility could be physically realized, i.e., where the matrix has to be generated and where it has to be pulled in or degraded. Finally, there is the possibility that the basal lamina is elastic and can stretch as suggested by Zamir et al. in their paper. This would have some analogy to fluid flow in a bounded domain as shown in Fig. 2A. Application of a localized force on an elastic membrane would result in a deformation that would generate two roughly counter-rotating sets of stress lines. If the basal lamina had elastoplastic properties, for instance as a result of local matrix remodeling, then it is conceivable in theory that this could result in apparent flows in the basal lamina. To get both forward and backward movements associated with bi-directional streak extension, as shown in our experiments (Chuai et al., 2006; Cui et al., 2005), it will be necessary to have two inward acting forces, which will result in compression followed by stretching in other areas (Fleury, 2005). In all these cases the question remains, which cells are moving actively and which cells are being displaced passively, which mechanisms drive their movements, and which signals coordinate their movements.

CONCLUSIONS

The fibronectin labeling experiments by Little and co-workers have provided some tantalizing results on the possible dynamics of the extracellular matrix during streak formation in the chick embryo. These results clearly demonstrate how little we know about the dynamics of in vivo cell-matrix interactions and what role they may play in large scale tissue movements. These results will need first of all confirmation by a series of independent experimental variations to test whether the labeled fibronectin molecules faithfully represent the dynamics of the matrix in vivo. If these results can be confirmed, then their thought provoking implications and some of the interpretations given by the Zamir et al. will need a lot more work to be substantiated and mechanistically understood. This will require the development and application of methods to identify cells that move actively and cells that are displaced passively, what forces are produced by cells in specific regions of the embryo, as well as methods to measure the mechanical characteristics of the extracellular matrix.