Active cell migration drives the unilateral movements of the anterior visceral endoderm

The early post-implantation mouse embryo consists of three tissues arranged like a cylinder, with extra-embryonic ectoderm positioned proximally, epiblast distally and visceral endoderm surrounding both(Fig. 1D). Cells of the epiblast give rise to the foetus whereas the other tissues contribute exclusively to extra-embryonic structures. The anterior visceral endoderm(AVE) is a specialised region of the visceral endoderm that, as much as a day before the formation of the primitive streak, expresses markers such as Hex (Hhex - Mouse Genome Informatics)(Thomas et al., 1998), Hesx1 (Hermesz et al.,1996; Thomas and Beddington,1996), Lim1 (Lhx1 - Mouse Genome Informatics)(Belo et al., 1997) and Cerberus-like (cerberus 1 homolog, Cer1 - Mouse Genome Informatics)(Belo et al., 1997; Shawlot et al., 1998).

The AVE plays an essential role in patterning the early embryo(Beddington and Robertson,1999). This was first demonstrated by microsurgical removal of the AVE, which causes loss of forebrain markers such as Hesx1(Thomas and Beddington, 1996). In addition, chimera experiments indicate that genes such as Hnf3b(Foxa2 - Mouse Genome Informatics)(Dufort et al., 1998), Otx2 (Rhinn et al.,1998), Smad2 (Madh2 - Mouse Genome Informatics)(Waldrip et al., 1998) and Lim1 (Shawlot et al.,1999) are required specifically in the visceral endoderm for proper patterning of the epiblast (reviewed by Martinez-Barbera and Beddington,2001). For example, embryos lacking Lim1 activity in the visceral endoderm completely lack anterior structures(Shawlot et al., 1999). The AVE functions in part by repressing the expression of posterior genes in the anterior epiblast (Kimura et al.,2000; Perea-Gomez et al.,2001), thereby causing the primitive streak to form on the opposite side of the embryo (Perea-Gomez et al., 2002). Although transplantation of the AVE alone is not sufficient to induce ectopic neural structures, it is capable of doing so in combination with anterior epiblast tissue(Tam and Steiner, 1999).

DiI-labelling experiments show that the AVE moves unilaterally from its initial position at the distal tip of the egg cylinder to the future anterior of the embryo, thereby converting a proximodistal axis to an anteroposterior axis (Thomas et al., 1998). Embryos mutant for cripto or Otx2 fail to effect this movement(Ding et al., 1998; Kimura et al., 2000; Perea-Gomez et al., 2001), and the epiblast becomes mispatterned such that the distal region adopts an anterior character and the proximal region a posterior character.

The central cells of the blastocoel floor of the Xenopus embryo(Jones et al., 1999) and the hypoblast of the chick embryo (Foley et al., 2000) are thought to correspond to the mouse AVE. The fates of these regions are similar to that of the AVE and, like the AVE, they express genes such as Hex. The regions also show functional similarities, capable of imposing anterior character on Xenopusectoderm (Jones et al., 1999)and restricting the formation of multiple primitive streaks in chick(Bertocchini and Stern, 2002). Significantly, the anterior movement of AVE cells is conserved in the equivalent tissues in both species (Jones et al., 1999), again highlighting the significance of this process.

Although the unilateral movement of AVE cells is crucial for anterior patterning of the embryo, very little is known about how it takes place. One suggestion is that it involves different rates of proliferation of anterior and posterior cells of the visceral endoderm or the polarised division of these cells (Lawson and Pedersen,1987; Thomas et al.,1998). Another is that the AVE might be carried unilaterally as part of a global movement of the entire visceral endoderm(Weber et al., 1999). In contrast to these passive means of movement, it is also possible that AVE cells actively migrate to their anterior position, implying that they respond to, or are directed by, environmental cues. It is impossible to distinguish between these possibilities by studying fixed specimens, and we have therefore developed a system to observe AVE movement in real time.

Our results show that AVE cells actively migrate from the distal tip of the egg cylinder to presumptive anterior regions. Time-lapse imaging reveals that this migration comes to an abrupt halt at the junction of the epiblast with the extra-embryonic ectoderm. Cell tracking reveals that once they reach this border, cells spread laterally in both directions, with highly convoluted paths. Confocal microscopy shows that migrating AVE cells retain direct contact with the epiblast at all times. Together, our results show that the anterior movement of AVE cells is the result of active cell migration, perhaps in response to cues from the epiblast or the extracellular matrix surrounding it.

Embryos carrying the Hex-GFP transgene were obtained from Hex-GFP mice (Rodriguez et al.,2001) maintained on a mixed CBA/J and C57BL6 background. Mice were kept on a 10-hour light, 14-hour dark cycle, and noon of the day of finding a vaginal plug was designated 0.5 dpc. Hex-GFP embryos were dissected at 5.25-5.75 dpc in M2 medium using forceps as described(Beddington, 1987), and tungsten needles were used to reflect Reichert's membrane. Embryos were cultured in an FCS2 chamber (Bioptechs) with a 0.5 mm spacer to create a culture cavity. The spacer was placed orthogonally to the intended orientation, so as to cover the unused perfusion ducts. The FCS2 chamber is designed to maintain the embryos at any desired temperature. The culture medium consisted of 50% DMEM/50% rat serum pre-equilibrated for 2 hours at 37°C, in an atmosphere containing 5% CO₂. Rat serum was prepared as described (Beddington,1987). Embryos were cultured for periods of up to 15 hours at a temperature of 37°C.

Phase contrast and epifluorescence digital time-lapse images were acquired using the Deltavision system from Applied Precision. Embryos were cultured directly on the stage of an Olympus IX70 inverted microscope and imaged using an Olympus 20× objective with a numerical aperture of 0.4. EGFP was excited using a standard 100 W mercury vapour lamp (Osram), with 490/20 excitation and 528/38 emission filters from Chroma. Exposure times were between 0.5 and 1.0 seconds per fluorescent image. Five images from different focal planes were usually captured at each time point. Fluorescent images from multiple focal planes were de-convolved and an extended-focus image was projected for each time point. When cultured embryos drifted in the field of view, projected images from different time points were manually set in register using Adobe Photoshop. QuickTime movies were compiled from individual still images using the Graphic Converter programme.

For the purpose of this study, filopodia were defined as cellular processes that were transient, were less than a fourth of a cell diameter in width at their base, and had at least one side that formed an angle of less than 120° with respect to the tangent that passed through the base of the filopodium. The Volocity program (Improvision) was used to outline 23 migrating AVE cells from seven embryos and then to calculate the centroids of each cell. The base and the tip of each filopodium on each cell were also marked, and their coordinates were determined. The length of each filopodium was calculated as the square root of[(x₂-x₁)²+(y₂-y₁)²],where (x₁, y₁) and (x₂, y₂) are the co-ordinates of the base and the tip of the filopodium. The `radius' of the cell was computed using the same formula, except that (x<sub>1</sub>,y<sub>1</sub>) and (x<sub>2</sub>, y<sub>2</sub>) were the coordinates of the base of the filopodium and the centroid of the cell. The length of the filopodium was expressed as a fraction of the `radius' of the cell. The angle of the filopodium with respect to the proximodistal axis of the embryo was determined as the arc tangent of[(y₂-y₁)/(x₂-x₁)], where(x₁, y₁) and (x₂, y₂) are the coordinates of the base and tip of the filopodium, respectively. The programme CricketGraph was used to plot the data as a polar graph.

Cell tracking was performed using the tracking module of the Volocity programme (Improvision). Eight cells were manually outlined using a graphics tablet and their centroids calculated at each time point. The path taken by cells was generated using Volocity. Data on the positions of the cells were imported into Microsoft Excel, where all further calculations were performed. The ratio of the movement of each cell was calculated as the absolute value of(y₂-y₁)/(x₂-x₁) where(x₁, y₁) and (x₂, y₂) are the coordinates of the centroids of a cell at two consecutive time points. The ratios of all eight cells tracked were then averaged for each time point. The distance separating two cells was calculated as the square root of[(x₂-x₁)²+(y₂-y₁)²]where (x₁, y₁) and (x₂, y₂) are the co-ordinates of the centroids of cells 1 and 2, respectively. The distance covered by a cell in each time interval was computed using the same formula,except that (x₁, y₁) and (x₂, y₂)were the coordinates of the same cell at two consecutive time points.

Embryos were fixed for 1 hour at 4°C in a solution of 4%paraformaldehyde in PBS, rinsed once at room temperature in PBT (0.1%Triton-100 in PBS) and then stained for 2 hours at 4°C in 0.5 μg/ml TRITC-Phalloidin (Sigma) in PBT. They were then washed once at room temperature in PBT and mounted on a slide using DAPI-Vectashield mounting medium (Vector Laboratories). Confocal images of the embryo were captured on a Leica TCS-SP upright microscope and de-convolved using the Hugyens programme from Scientific Volume Imaging. Confocal stacks were rendered as 3D volumes using Volocity (Improvision).

Embryos were dissected at roughly 5.75 dpc, after the AVE was likely to have moved anteriorly. Embryos were fixed in a solution of 4% paraformaldehyde in PBS and in situ hybridisation was carried out following standard procedures(Wilkinson, 1992). Hex (Thomas et al.,1998), and Cer1(Thomas et al., 1997) probes were as described.

Embryos at the early headfold stage and later can be maintained in rolling culture, under which conditions they show remarkably normal development(Tam, 1998). More recently it has become possible to maintain such embryos in static culture, making it possible to image their development in real time(Jones et al., 2002). Embryos at earlier, pre-streak stages also develop well in static culture, even progressing as far as the primitive streak stage(Thomas et al., 1998). In order to image such embryos, we adapted a commercially available culture chamber intended for imaging living cells (see Materials and methods). Though DiI and other vital dyes have long been used to label cells during embryogenesis, there is evidence that in early mouse embryos the invasive labelling of cells perturbs their behaviour by delaying cell division(Piotrowska et al., 2001). Therefore, to visualise the cells of the AVE, we used transgenic mice that express enhanced green fluorescent protein (EGFP) under the control of Hex regulatory sequences(Rodriguez et al., 2001). These embryos recapitulate the early expression pattern of Hex(Thomas et al., 1998),initially in the distal visceral endoderm and later in the AVE.

Embryos were dissected and set up in culture at approximately 5.5 dpc(early egg cylinder stage), when Hex is expressed at the distal tip of the embryo. Phase-contrast and EGFP fluorescence images were captured every 10-15 minutes over a period of 8-10 hours. Of 32 embryos cultured under these conditions, 27 developed normally, as indicated by morphological criteria(expansion of pro-amniotic cavity into the extra-embryonic ectoderm and overall growth) and by the unilateral shift of the AVE from the distal tip of the egg cylinder to the prospective anterior(Fig. 1, and see Movie 1).

In 12 of the 27 embryos that developed normally, we were able to discern individual cells of the AVE in great detail as they moved to the prospective anterior of the embryo. AVE cells at the distal tip of the embryo are columnar, tightly clumped together and inactive (Figs 2, 5). As soon as movement is initiated, however, they undergo a dramatic change in morphology. In particular, they become squamous and motile, and they project filopodial processes, primarily in the direction of motion(Fig. 2, Fig. 3A-D,G, and Movies 3 and 5). Filopodia are frequently greater than one cell radius in length(Fig. 3G) and they often make contact with surrounding cells (Fig. 3A-D), allowing for the possibility of intercellular communication. AVE cells are occasionally seen to divide(Fig. 3E,F), but not in any consistent orientation. They show the hallmarks of migration, in that they translocate and project filopodia, predominantly in the direction of their motion.

AVE cells generally traverse the approximate 100 μm from the distal tip of the egg cylinder to the prospective anterior in 4-5 hours (Figs 1, 2). On reaching the junction of the epiblast and the extra-embryonic ectoderm they come to an abrupt halt(Figs 1, 2), as if encountering a physical barrier (see Movies 2 and 4). The cells then start to spread laterally on the epiblast, they no longer send out filopodia and they become elongated, with their long axes parallel to the junction of the epiblast and extra-embryonic ectoderm(Fig. 2).

To compare the trajectories of individual AVE cells during migration, we computed the positions over time of AVE cells in representative embryos. Fig. 4 shows the tracks of eight cells in the embryo shown in Fig. 2 (see Movies 2 and 3). The tracks of AVE cells in two other embryos (Movies 4, 5, 6 and 7) are shown in supplementary Fig. S1. The tracks of the eight cells in Fig. 4 show that not all reach the extra-embryonic ectoderm. The leading cells (cells 1, 2 and 3 in Fig. 4A) do reach the extra-embryonic ectoderm and start to spread laterally. The cells behind them,however (cells 4, 5, 6, 7 and 8 in Fig. 4A,B), stop migrating proximally and start spreading laterally before reaching the extra-embryonic ectoderm, presumably because they are obstructed by the leading cells. Cells 5 and 6, which are sisters, share a common track before they divide. They have separate tracks after division, but remain in contact with each other throughout their further migration(Fig. 4B). The same is observed for cells 4 and 7, which are also sister cells(Fig. 4A,B).

When migrating proximally, all eight cells show a rather direct trajectory,but when spreading laterally their trajectories become highly convoluted,albeit with a net lateral movement. This can be illustrated by calculating the ratios of the proximal movements of cells to their lateral movements at different times during development (Fig. 4C). Initially, the means of these values are greater than unity,indicating that AVE cells tend to move proximally. However, when the leading cells reach the boundary of the extra-embryonic ectoderm (red arrow, Fig. 4C) the means fall below one, indicating that although their tracks become convoluted the net displacement of AVE cells is lateral rather than proximal.

The tracks of all cells in Fig. 4 move to the left at the boundary. This is a result of the embryo`rolling' during culture (see Movie 2 at 7 hours of culture). To correct for this rolling, and to compare the lateral directions in which cells move at the boundary, we calculated the distances between cells over time, which should be less affected by the rolling. The separation between two representative pairs of cells is shown in Fig. 4D. Cells 1 and 3 move away from each other after reaching the boundary, whereas cells 2 and 4 move towards one another. This indicates that on reaching the boundary, the lateral movement of AVE cells does not occur in a coordinated manner, but that cells move independently of one another in either direction.

Our observations show that AVE cells migrate from the distal region of the embryo to the junction between the epiblast and the extra-embryonic ectoderm. Once AVE cells have migrated some distance proximally, EGFP-expressing cells can be seen intermingled with non-expressing cells (Figs 2, 3). To investigate the significance of this apparent mixing and to determine whether AVE cells migrate on top of more proximal visceral endoderm cells, we examined phalloidin-stained prestreak Hex-GFP transgenic embryos by confocal microscopy.

The AVE is often discernible as a thickening of the visceral endoderm(Fig. 5A). Kimura and colleagues have suggested that the AVE is thicker than surrounding visceral endoderm because it consists of two cell layers that form a stratified cuboidal epithelium (Kimura et al.,2000). However, detailed confocal imaging of six embryos shows that the AVE comprises a single layer of approximately 10-15 cells, their columnar nature accounting for the observed thickening(Fig. 5). The columnar cells of the AVE are polarised, their nuclei displaced towards the epiblast and showing stronger actin staining on the surface away from the epiblast(Fig. 5C).

As AVE cells migrate proximally, they lose their columnar nature and become squamous. In confocal sections through nine embryos, migrating AVE cells are never seen on top of other visceral endoderm cells but are apposed to cells of the epiblast (Fig. 5D). This suggests that anterior migration occurs directly on the epiblast rather than on other cells of the visceral endoderm.

Significantly, we observe that even prior to migration, when AVE cells are at the distal tip of the embryo, they comprise a population of both Hex-GFP expressing and non-expressing cells(Fig. 5C), indicating that the salt-and-pepper appearance of migrating AVE cells is likely to be a consequence of the initial heterogeneity within the AVE. Although some intermingling with non-AVE cells may occur during migration, this is likely to be limited, because little cell mixing occurs in the visceral endoderm(Gardner and Cockroft, 1998). A similar salt-and-pepper pattern of expression is seen with endogenous Hex transcripts as well as other markers of the AVE, such as Cerberus-like (Fig. 6),suggesting that expression of the Hex-GFP transgene is a true reflection of Hex transcription and not a result of position effect variegation.

We followed the behaviour of cells of the anterior visceral endoderm (AVE)of the mouse embryo as they moved towards the proximal anterior region of the embryo. Confocal microscopy revealed that the AVE consists initially of a single layer of cells that is tightly packed, polarised and columnar, causing a discernible thickening compared with the rest of the visceral endoderm(Fig. 5A-C). The expression of Hex at this stage, as revealed by a Hex-GFP transgene, is heterogeneous, with some cells expressing the gene and others not. This salt-and-pepper pattern is maintained within the AVE during migration, with GFP-expressing cells intermingled with non-expressing cells. In situ hybridisation experiments confirmed that endogenous Hex transcripts are also expressed in a heterogeneous manner. This suggests that the AVE comprises several different cell types that express different molecular markers. A detailed and high-resolution analysis of the expression domains of such AVE markers will be needed to resolve this issue.

As they begin to move proximally, cells of the AVE become squamous and their behaviour becomes highly dynamic; they change their shapes continuously and project filopodia in the direction in which they are moving. Our results suggest that most of the movement of AVE cells can be accounted for by migration. The increase in the surface area occupied by AVE cells as a result of their becoming squamous might contribute to their translocation to some extent, but is unlikely to be the primary motive force. For example, cells occasionally switch positions with respect to one another (cells 3 and 4 in Fig. 4A and Movie 3; cells 2 and 3 in supplementary Fig. S1, panel B, and Movie 7). This would not occur if cell movement were the result of passive expansion due to changes in cell shape. Cell division is also unlikely to contribute significantly to their movement because cleavage does not occur with a consistent orientation and because the movement is completed in four to five hours, during which time only one or two AVE cells are usually observed to divide. The sister cells in Fig. 4B, for example, have migrated a substantial distance proximally before they divide. Global movement of the surrounding visceral endoderm is also unlikely to contribute significantly to AVE movement. Although most AVE cells move proximally, some invariably remain at the distal tip of the embryo (Figs 1, 2). This would not occur if the entire visceral endoderm were moving.

Our data show that AVE cells migrate proximally, to the boundary of the epiblast and extra-embryonic ectoderm, and then abruptly begin to move laterally. This boundary provides an endpoint to the proximal migration of AVE cells and positions them such that they can undertake their subsequent role of patterning the underlying epiblast. We note that not all migrating AVE cells reach the boundary; rather, those that arrive first appear to prevent cells behind them from moving further proximally. It is possible that cells at the boundary become `squashed' by later-arriving cells, causing them to become elongated along an axis parallel to the junction between the epiblast and extra-embryonic ectoderm.

In retrospect, the existence of this boundary can be inferred from previous cell lineage analyses. When pre-streak embryos are labelled in the region we now recognise as the AVE, and then cultured to early streak stages, labelled cells are observed to have moved proximally, but never beyond the epiblast(Lawson and Pedersen,1987).

What prevents leading AVE cells from migrating beyond the epiblast? One possibility is that migration is controlled by the underlying epiblast. Confocal images show that AVE cells migrate in direct contact with the epiblast. The functions of genes such as cripto and β-catenin, which are necessary for AVE migration, are required in the epiblast and not the visceral endoderm (Ding et al., 1998; Huelsken et al., 2000). If AVE cells require contact with the epiblast in order to migrate, this would explain why they do not migrate onto the extra-embryonic ectoderm.

Alternatively, signals from the extra-embryonic ectoderm (or the visceral endoderm overlying the extra-embryonic ectoderm) might actively repel AVE cells, preventing them from migrating beyond the epiblast. We note that later in development AVE cells do move beyond the epiblast; during gastrulation they are displaced onto the forming yolk sac by the anterior definitive endoderm(Lawson and Pedersen, 1987; Thomas and Beddington, 1996). This might occur because AVE cells have themselves changed, as suggested by the fact that they downregulate markers such as Hex(Thomas et al., 1998). Another possibility is that by these stages the extra-embryonic ectoderm has been displaced by the forming yolk sac, which does not repel AVE cells. Interestingly, in embryos mutant for angiomotin, the AVE is not displaced onto the extra-embryonic region during gastrulation(Shimono and Behringer, 2003). Analysis of these mutant embryos should help resolve this issue. We hope to address these, and other questions, by continuing to observe mouse embryos in real time as they develop in culture.

This work was supported by the Medical Research Council and the Wellcome Trust. We thank Stamatis Pagakis for help with imaging and Jean-Baptiste Manneville for useful discussions about cell tracking. Frank Costantini, Val Wilson and Irene Yan provided helpful comments on the manuscript. S.S. and T.R. would like to thank Lyle Zimmerman for the use of his lab space. S.S. is a fellow of the Human Frontier Science Program.