Coordination of cell proliferation and anterior-posterior axis establishment in the mouse embryo

The anterior visceral endoderm (AVE) is an extra-embryonic signalling centre that establishes the anterior-posterior (A-P) axis of the mouse embryo. The AVE forms at the distal tip of the embryo at 5.25 days post-coitum (dpc) by a combination of inductive signalling by the Tgfβ factors Nodal (Brennan et al., 2001) and activin (Yamamoto et al., 2009) and inhibitory signalling by the extra-embryonic ectoderm (Rodriguez et al., 2005; Yamamoto et al., 2009). At 5.5 dpc, cells from the AVE undergo a unidirectional movement from the distal tip of the embryo to the future anterior side. This movement is essential for correct positioning of the A-P axis (Srinivas, 2006; Takaoka et al., 2007).

Although the precise mechanism of how cells of the AVE move from the distal tip [where they are termed distal visceral endoderm (DVE)] to the prospective anterior is not well understood, two alternative hypotheses have been proposed. Time-lapse imaging studies have suggested that AVE movements are part of an active migration process (Srinivas et al., 2004). The findings that Nap1 (Nckap1 – Mouse Genome Informatics; a component of the WAVE complex that is essential for cell migration) and the small GTPase Rac1 are required for AVE movements (Migeotte et al., 2010; Rakeman and Anderson, 2006) and that canonical Wnt signalling and its antagonists act as guidance cues for AVE migration (Kimura-Yoshida et al., 2005) have supported this hypothesis. In contrast to these findings, BrdU-incorporation studies coupled with embryo transfection experiments have led to the proposal that differential proliferation within the VE, modulated by Nodal signalling, provides the driving force for AVE movements. According to this model, Nodal signalling is required for proliferation within the VE, and prospective AVE cells are displaced because of the differential rate of proliferation that is created when Nodal activity is antagonised only in the anterior of the embryo by Lefty1 and Cerberus-like (Cer1 or Cerl1) (Yamamoto et al., 2004). However, further experiments are required to clarify the relative contributions of both these processes to the movements of the AVE.

Concurrent with AVE movements, the underlying epiblast undergoes rapid proliferation. It has been calculated that there is a 4.5- to 5-fold increase in epiblast cell number between 5.5 and 6.5 dpc, with the epiblast requiring ~660 cells for gastrulation to commence (Snow, 1977). Formation of the primitive streak at the prospective posterior of the embryo at 6.5 dpc marks the initiation of gastrulation and this dramatic rise in epiblast cell number is required to sustain the recruitment of ectoderm cells through the primitive streak. This suggests coordination between epiblast growth and posterior patterning of the embryo.

What factors are responsible for maintaining proliferation in the early embryo? Nodal mutant embryos are smaller than controls (Camus et al., 2006; Mesnard et al., 2006) and it has been reported that Nodal promotes cell proliferation in the VE (Yamamoto et al., 2004). Bmpr1a mutant embryos show epiblast proliferation defects (Mishina et al., 1995), implicating BMP signalling in the maintenance of proliferation in the early post-implantation embryo. Indeed, BMP signalling is required to sustain Nodal signalling in the epiblast (Ben-Haim et al., 2006; Di-Gregorio et al., 2007), suggesting that these two signalling pathways might cooperate to sustain epiblast proliferation.

To address the impact of proliferation on A-P axis establishment, we have carried out a detailed analysis of the proliferation patterns within the VE and epiblast at the time of AVE movements. We have studied how this proliferation is regulated and tested its importance for the establishment of the A-P axis in the mouse embryo.

Mice carrying the Hex-GFP transgene (Rodriguez et al., 2001) and the CAG-geminin-GFP reporter transgene (Sakaue-Sawano et al., 2008) have been described previously. Embryos were dissected in M2 as described (Nagy et al., 2003). Mice were maintained and treated in accordance with the Home Office's Animals (Scientific Procedures) Act 1986.

Hex-GFP embryos were sorted for the presence of GFP and fixed in 4% paraformaldehyde overnight at 4°C. Standard protocols for immunohistochemistry (Nagy et al., 2003) were used for rabbit anti-phospho-histone H3 (Ser28) (Upstate, 07-145; diluted 1:500), Pcna (Abcam; 1:500) and Ki67 (Abcam; 1:500) and for WISH (Thomas and Beddington, 1996). A sodium citrate antigen-retrieval step was carried out for Ki67 and Pcna. Stained embryos were placed in glass-bottom microwell dishes (MatTek, USA) in a drop of 1:1 glycerol:PBS. Confocal images of the embryo were captured on a Leica DM IRB inverted confocal microscope through either a 10×/0.30 PH1 air objective or a 40×/1.00 PH3 oil-immersion objective. z-stacks were taken at 4-μm intervals through the embryo, with each channel acquired sequentially. Stacks were then merged and reconstructed using ImageJ 1.33u.

A modification of the protocol available on the Rossant Laboratory web page (www.sickkids.ca/research/rossant/protocols/BrdU_embryoStain.asp) was used. Briefly, Hex-GFP embryos were cultured in 20 μM BrdU in 80 μl drops of 1:1 DMEM:fetal calf serum covered by mineral oil at 37°C, 5% CO₂ for 15 or 30 minutes, washed 3× 5 minutes in PBS and fixed in 4% paraformaldehyde containing 2 mM EGTA for 1.5 hours at 4°C. Embryos were then washed in PBT (PBS containing 0.1% Tween 20) for 1 hour at 4°C and blocked in PBS containing 10% normal goat serum and 0.1% Triton X-100 at room temperature (RT) for 1.5 hours. Embryos were then incubated with rabbit anti-GFP antibody (Molecular Probes, A11122) diluted 1:1000 in blocking solution at 4°C for a minimum of 2 days. Embryos were washed 3× 5 minutes plus 1× 1 hour at RT in 0.1% Triton X-100 and incubated in Alexa Fluor 488 goat anti-rabbit IgG (Molecular Probes, A11008) diluted to 1:500 with blocking solution for 2 hours at RT with rocking. Embryos were washed 3× 5 minutes plus 1× 1 hour at RT, post-fixed in 4% paraformaldehyde for 10 minutes and transferred to 2 M HCl in PBS containing 0.1% Tween 20 for 30 minutes at 37°C to denature the DNA. The HCl was quenched by washing 3× 5 minutes in borate buffer (0.1 mM, pH 8.5) at RT, blocked in MABT (100 mM maleic acid pH 7.5, 150 mM NaCl, 0.1% Tween 20) containing 10% sheep serum for a minimum of 1.5 hours at RT. The embryos were then incubated overnight in rat anti-BrdU (Abcam, ab6326) diluted 1:100 in MABT containing 1% sheep serum at RT with constant rocking. The following day, embryos were washed 3× 5 minutes in MABT containing 10% sheep serum and 5× 1 hour in MABT at RT. Embryos were incubated in polyclonal Cy3 goat anti-rat IgG (Abcam, ab6953) diluted 1:500 with the same blocking solution and incubated at 4°C overnight. Finally, embryos were washed 5× 1 hour in MABT and stained with To-pro-3 iodide (Molecular Probes, T3605) diluted 1:500, mounted and images acquired.

Hex-GFP embryos were cultured in 200 nM nocodazole (Sigma) (MacAuley et al., 1993; Vasquez et al., 1997), 200 nM nocodazole plus 500 μg/ml Dkk1 (R&D Systems), 150 μM genistein (DiPaola, 2002), 2.5 μM hydroxyurea (Engstrom et al., 1979) (Sigma) or 10 μM SB-431542 (Inman et al., 2002) in pre-equilibrated drops. Each 80 μl drop comprised inhibitor diluted in 1:1 DMEM:rat serum covered with mineral oil. Embryos were incubated at 37°C, 5% CO₂ overnight (typically 16 hours) or for 2 hours (see Fig. S4 in the supplementary material), then washed 3× 5 minutes in PBS and fixed in 4% paraformaldehyde overnight at 4°C. Embryos then underwent staining as described above. In the case of BrdU treatment of embryos cultured in SB-431542, 20 μM BrdU was added for the last 30 minutes of culture and then embryos were stained as described above.

Cells were scored using the ImageJ cell counter. The non-parametrical χ² test was used in Fig. 2 to test the difference between the sample (number of dividing cells in each quadrant) and expected distribution if proliferation were similar in all regions of the embryo (each quadrant is assumed to represent 25% of dividing cells). t-tests were applied to the data in the remaining figures. This parametric test is based on the assumption that the samples come from populations that are normally distributed using continuous, not discrete, numerical data. The unpaired t-test assumes that the two data sets being compared are independent and identically distributed (i.e. from different embryos but both normally distributed).

As a starting point to study whether differential proliferation could be the driving force for AVE movements, time-lapse imaging of Hex-GFP embryos (which have a fluorescently labelled AVE) was carried out (Rodriguez et al., 2001). The positions of GFP-positive cells at the leading (anterior) and lagging (posterior) fronts of the Hex-GFP domain were analysed at the onset of AVE movements. As the anterior AVE cells moved proximally, some GFP-positive cells remained distal (Fig. 1A) and did not migrate in the time frame that it took the leading cells to reach the embryonic/extra-embryonic boundary. This result suggests that at least the leading AVE cells are not passively displaced to the anterior of the embryo by increased proliferation in the GFP-negative posterior part of the VE.

We set out to examine in greater detail the proliferation rates of different regions of the VE during the time that AVE movements are occurring. As a first step, we determined the overall levels of proliferation in the anterior and posterior VE by analysing the distribution of Ki67-positive cells, a protein that is expressed in all proliferating cells (Chenn and Walsh, 2002). At 5.75 dpc, 97.5±1.18% of Hex-GFP AVE cells were Ki67 positive, compared with 94.2±2.59% of posterior VE cells. At 6.25 dpc, 99.46±0.61% of AVE cells were labelled, as compared with 98.72±0.81% of posterior VE cells (see Fig. S1 in the supplementary material). This indicates that very few cells in any region of the VE are not in the cell cycle.

To analyse the distribution of cells in S phase we studied the pattern of BrdU incorporation in the VE of 5.5-dpc embryos. Hex-GFP embryos were given a 15-minute BrdU pulse and then divided into three groups according to GFP expression: the pre-distal group displayed a low level of GFP expression throughout the VE (Fig. 1B); the distal group included those embryos in which the DVE had resolved to the distal tip (Fig. 1D); and the tilted group showed a clear asymmetric tilt of the DVE (Fig. 1F). The numbers of BrdU-positive nuclei were counted and expressed as a percentage of the total cell number, referred to as the incorporation index. In the pre-distal group, the mean incorporation index of the VE was 76±2.66% (Fig. 1C), with non-BrdU-incorporating nuclei not appearing to cluster in any particular region of the VE. Regional division of the embryonic VE was not possible at this stage because GFP expression was throughout the VE (Fig. 1B).

Although an A-P axis could not be assigned to the distal group, embryos were divided into DVE, embryonic VE (embVE) and extra-embryonic VE (exeVE) domains (Fig. 1D). The mean incorporation index of these areas was quantified and found to be very similar: DVE, 68.38±4.81%; embVE, 70.07±4.91%; and exeVE, 72.63±4.97% (Fig. 1E).

The tilted group could be assigned an A-P axis and the embVE was divided into anterior and posterior halves according to position relative to the DVE (Hex-GFP domain). There was no significant difference in the mean incorporation index of the anterior (76.54±2.68) and posterior (73.84±4.74%) VE (Fig. 1G). To ascertain if there were any regional differences within the anterior VE, this region was further subdivided into VE anterior to the DVE and DVE. Again, both these areas showed very similar incorporation indexes: VE anterior to the DVE, 80.78±3.18%; DVE, 74.44±3.63% (see Fig. S2A,B in the supplementary material). Next, the ratio of the anterior to posterior VE proliferation index was plotted for each embryo to analyse the variability that might exist between embryos. A value exceeding 1 indicated more proliferation in the anterior than posterior VE, whereas a value of less than 1 indicated the converse. Of the 15 embryos in the group, eight displayed values exceeding 1 and seven had values below 1 (Fig. 1H), which is inconsistent with there being increased proliferation in the posterior as compared with the anterior VE. Therefore, when the DVE movements are being initiated, the distribution of cells in S phase is not significantly different between the anterior and posterior VE.

To confirm that our BrdU data accurately reflect the distribution of proliferating cells, we analysed the distribution of cells in M phase within the VE between 5.25 and 5.75 dpc using an anti-phospho-histone-H3 (PH3) antibody. Embryos were grouped as before according to the position of the Hex-GFP domain (Fig. 2A), and the mitotic index (percentage of mitotic cells; see Table S1 in the supplementary material) was calculated for the different regions of the VE. In the distal group there was no significant difference in the mitotic index of the DVE (2.34±1.71%), embVE (1.33±0.99%) and exeVE (1.63±0.89%; Fig. 2B). In the tilted group the mean mitotic index was similar in the anterior and posterior quadrants (1.44±0.62% and 1.43±0.66%), which were higher, although not significantly so, than those of the lateral quadrants (right, 1.01±0.65%; left, 0.56±0.55%).

To rule out the possibility that a proliferative burst within the VE might have been missed in the previous experiment, embryos were cultured for 4 hours in the presence of the mitotic inhibitor nocodazole (MacAuley et al., 1993; Vasquez et al., 1997) just as DVE movements are being initiated. This caused a 3-fold accumulation in the proportion of mitotic cells (Fig. 2D,E). When the mitotic index was calculated for each quadrant in these nocodazole-treated embryos (Fig. 2F), there was no significant difference between the anterior (9.89±3.08%) and posterior (7.01±2.86%) quadrants. To test whether there were any regional differences within the anterior VE, this region was further subdivided into VE anterior to the DVE and DVE, as described above. We again found that the mitotic index of these anterior regions was not significantly lower than that of the posterior VE: VE anterior to the DVE, 11.82±5.45%; DVE, 6.15±1.66% (see Fig. S2C,D in the supplementary material). By contrast, when we compared the lateral quadrants we found that the mean mitotic index of the lateral right quadrant was significantly higher (12.36±1.54%) than that of the lateral left quadrant (4.89±2.16%; P<0.01), suggesting that proliferation on the right side of the embryo is greater than on the left. In conclusion, we observe little difference between the proliferation rates of the anterior and posterior VE and it is therefore unlikely that the first cells of the AVE initiate their movements by being displaced as a result of differential proliferation rates.

It is possible that differential proliferation within the VE could be responsible for the movement of AVE cells that do not migrate with the leading edge of the AVE at 5.5 dpc. To test this hypothesis, 6.25-dpc embryos were labelled with BrdU for 30 minutes and the incorporation index was calculated (Fig. 1I,J). The AVE (all anterior VE cells) had a mean incorporation index of 61.44±5.25%, a similar value to the 64.13±4.61% obtained for the rest of the VE (Fig. 1J), suggesting no A-P difference in proliferation.

To further assess the regional distribution of mitotic cells within the embryonic VE at 6.25 dpc, PH3-stained embryos were analysed (Fig. 3A,B). The anterior quadrant had a mitotic index of 1.81±0.82%, the posterior quadrant 1.59±0.67%, the lateral left quadrant 1.33±0.65% and the lateral right quadrant 1.66±1.12%. This small variation in mitotic index between quadrants was not statistically significant, suggesting that the low levels of cell division within the VE at 6.25 dpc show no regional bias. However, we cannot exclude the possibility that local proliferation differences do exist within the subdomains analysed.

To determine whether there was any difference between the cell cycle rates of the AVE and the rest of the VE, 6.25-dpc Hex-GFP embryos were cultured in nocodazole for intervals of 0-5 hours (Fig. 3C). The mitotic index of the VE was calculated for each embryo at each time point and the percentage of cells in mitosis was plotted as a function of time in the presence of nocodazole (Fig. 3D). There was no significant difference between the division rates of the anterior and posterior VE at any of the time points studied, indicating that cells of the AVE cycle at a similar rate to those in the rest of the VE.

This random distribution of cells in S phase and in M phase contrasts with a previous report of no BrdU incorporation in the AVE at 6.5 dpc (Yamamoto et al., 2004). We therefore confirmed that there were no significant differences between the proliferation rates of cells in the AVE and posterior VE by analysing the distribution of cells positive for Pcna (a marker of cells in G1-S) (Connolly and Bogdanffy, 1993; Takahashi and Caviness, 1993). At 6.25 dpc we observed very similar levels of Pcna-positive cells between the anterior and posterior VE (89.00±2.83% and 83.85±4.70%, respectively; see Fig. S3 in the supplementary material). We also analysed the distribution of geminin-positive cells using a transgene carrying a CAG-geminin-GFP reporter. This transgene shows highest GFP expression during the late S phase and G2 phases of the cell cycle (Sakaue-Sawano et al., 2008). We found no significant difference in the numbers of GFP-positive cells between the anterior and posterior VE (12.77±2.24% and 14.08±3.17%, respectively; see Fig. S3 in the supplementary material) of 6.25-dpc embryos using this reporter. Together, these results argue against the hypothesis that, at any time point, migration of AVE cells is driven by differential proliferation within the VE.

To study how proliferation rates change within the epiblast as AVE movements are initiated, we analysed PH3-stained Hex-GFP embryos and determined the mitotic index for this tissue in pre-distal, distal and tilted group embryos (Fig. 4A). We found little variation in the mitotic index of the epiblast between these stages (pre-distal group, 8.05±1.31%; distal group, 8.68±1.43%; tilted group, 7.96±1.23%). This contrasts with a significant decrease in mitotic levels within the VE during this period (3.65±1.17%, 1.56±0.68%, 0.72±0.29%, respectively; P<0.02). These data indicate that whereas the proliferation rate of the epiblast is 2.2-fold that of the VE prior to AVE induction, when AVE movements are initiated it is 11-fold that of the VE, creating a severe growth imbalance between these tissues. By 6.25 dpc, the overall level of proliferation in the epiblast had dropped and the difference in proliferation rates between the epiblast and VE had fallen to a value similar to that prior to AVE migration (3.7-fold; epiblast, 5.74±0.84%; VE, 1.56±0.66%; Fig. 4A).

The regional distribution of mitotic nuclei in the epiblast at 5.5 dpc was examined (Fig. 4B). When the mean mitotic indexes of the anterior and posterior epiblast of tilted group embryos were compared, no significant difference was observed (anterior, 7.13±1.24%; posterior, 8.31±1.57%). The proliferation rates of the proximal (9.06±1.50%) and distal (5.36±1.27%) epiblast were also not significantly different (Fig. 4B). Finally, the right quarter of the epiblast had a significantly greater mitotic index (5.06±0.57%) than the left quarter of the epiblast (2.41±0.63%; P<0.01; Fig. 4B). This observation, combined with the fact that a similar trend is seen within the VE (Fig. 2F), indicates a higher proliferation rate on the right side of the embryo at 5.5 dpc.

We repeated this analysis at 6.25 dpc (Fig. 4C) and found no significant difference between the mitotic indexes of the anterior (6.46±0.75%) and posterior (6.61±0.55%) epiblast or between the left (2.93±1.05%) and right (4.02±0.62%) epiblast. However, the proximal epiblast had a significantly higher mitotic index (8.01±1.01%) than the distal epiblast (3.67±0.61%; P<0.005), which correlates with this being the site of initiation of primitive streak formation, a tissue that will subsequently exhibit a higher rate of proliferation (MacAuley et al., 1993; Snow, 1977).

Nodal signalling has been shown to be required for the movements of AVE cells (Norris et al., 2002) by driving differential proliferation within the VE (Yamamoto et al., 2004). Given that we observe no evidence for differential proliferation within the VE at the time of AVE movements, we revisited the requirements for Nodal in these movements. Hex-GFP embryos at 5.5 dpc were cultured overnight in the presence of SB-431542, a small-molecule inhibitor of the Nodal receptor (Inman et al., 2002). This resulted in a complete lack of AVE migration (Fig. 5A′), confirming that Nodal signalling is required for AVE movements. Proliferation was determined in these embryos by BrdU incorporation. Once again, embryos were subdivided into three regions: the GFP-positive AVE, the GFP-negative VE and the epiblast (Fig. 5A-B). In control embryos, the mean BrdU-incorporation index of the AVE (46.23±2.49%) was similar to that of the rest of the VE (50.85±3.47%) and lower than that of the epiblast (66.26±3.25%), reflecting the shorter cell cycle length in the epiblast. Interestingly, after inhibition of Nodal signalling there was little change in the BrdU-incorporation indexes of the AVE (43.97±3.17%) and the rest of the VE (53.12±2.60%), but the incorporation rate of the epiblast had dropped dramatically when compared with controls (8.78±2.43%; P<0.0001). This indicates that whereas proliferation in the epiblast is Nodal dependent, cell division within the VE is Nodal independent. To confirm that proliferation in the epiblast is indeed reduced after inhibition of Nodal signalling, PH3 staining was carried out on SB-431542-treated embryos (Fig. 5C,D). Blocking Nodal signalling caused a significant reduction in the number of mitotic cells in the epiblast, with control embryos having a mean mitotic index of 7.85±0.64% and SB-431542-treated embryos a mean mitotic index of 4.21±0.4% (P<0.0001). The possibility that epiblast cells exit from the cell cycle after SB-431542 treatment was supported by the observation that expression of c-Myc was also lost from the epiblast after inhibition of Nodal signalling (see Fig. S4A′ in the supplementary material).

Nodal signalling is required to induce the AVE (Brennan et al., 2001), so one explanation for the lack of AVE migration after inhibition of Nodal signalling is a failure to maintain the identity of this tissue. The expression of the AVE markers Dkk1, Lim1 (Lhx1 – Mouse Genome Informatics) and Cerl1 and the proximal VE markers Hnf4 and Gata4 were analysed in 5.5-dpc embryos after treatment with SB-431542. There was a loss of expression of all the AVE markers (see Fig. S4B-D′ in the supplementary material), an expansion in Hnf4 expression (see Fig. S4F′ in the supplementary material) and a strong downregulation in the expression of Gata4, as compared with control embryos (see Fig. S4E′ in the supplementary material). The loss of expression of AVE markers and the upregulation of Hnf4 in the distal VE indicate that Nodal signalling is required not only for AVE induction but also for the maintenance of AVE identity in these cells. Therefore, Nodal signalling is required for the maintenance of proliferation in the epiblast and for expression of AVE markers in the VE.

The misspecification of AVE markers is likely to be one cause for the AVE migration defects observed after inhibition of Nodal signalling, but this does not exclude the possibility that Nodal-driven proliferation in the epiblast is also required for AVE migration. To test this possibility, we studied embryos mutant for the Nodal co-receptor Cripto (Tdgf1 – Mouse Genome Informatics), which specify a DVE, but the DVE does not migrate to the prospective anterior (Fig. 5E,F) (Ding et al., 1998). The mitotic index of the VE of Cripto-null embryos (2.92±0.92%) was similar to that of the controls (2.13±0.5%); however, in the epiblast, the mitotic index of Cripto mutant embryos (3.76±0.38%) was significantly lower than that of controls (6.75±0.82%; P<0.05). Given that Cripto is required in the epiblast for the migration of the AVE (Kimura et al., 2001), this suggests that this defect in epiblast proliferation could contribute to the migration defects observed in Cripto mutant embryos.

To confirm this result, we cultured 5.5-dpc embryos for 2 hours with the Nodal receptor inhibitor SB-431542 when AVE movements are just being initiated. During the first phase of inhibition of Nodal signalling, we observed no change in the expression of AVE markers, nor in the expression of general VE markers (see Fig. S5A-D′ in the supplementary material). By contrast, we observed a significant decrease in the proliferation rate of the epiblast but not of the VE (see Fig. S5E-F in the supplementary material). These data suggest that during the first 2 hours of Nodal signalling inhibition, the primary process that is affected is proliferation rather than AVE specification. As it is during this period that the AVE movements are initiated, these results show that Nodal-driven proliferation in the epiblast affects AVE migration.

To test whether proliferation in the epiblast is required for AVE migration, cell cycle inhibitors were used. Hex-GFP embryos at 5.5 dpc were cultured overnight in the presence of the mitotic inhibitor nocodazole (200 nM) (MacAuley et al., 1993). This resulted in a block in AVE migration (n=16/18; data not shown) and in at least a 3-fold accumulation of cells in M phase (Fig. 2E), suggesting that proliferation is indeed required for AVE migration. To confirm these results, two other cell cycle inhibitors were tested: hydroxyurea, which blocks cells in S phase (Engstrom et al., 1979), and genistein, a tyrosine kinase inhibitor that blocks the G₂-M transition (Matsukawa et al., 1993). Embryos cultured in 2.5 μM hydroxyurea showed a lack of AVE migration (n=15/17; Fig. 6A′) and displayed a 93.6% reduction in the number of mitotic cells in the epiblast (controls, 10.00±0.73%; hydroxyurea treated, 0.64±0.39%; P<0.0001; Fig. 6B). There was no significant difference in the number of mitotic cells in the VE of these embryos (controls, 1.51±0.96%; hydroxyurea treated, 0.67±0.67%; Fig. 6B). Similarly, embryos cultured in 150 μM genistein also presented a failure of AVE migration (n=7/11; Fig. 6C,C′) and a significant decrease in mitotic cells in the epiblast (controls, 10.54±0.75%; genistein treated, 4.51±1.1%; Fig. 6D; P<0.0005) but not within the VE (controls, 3.02±0.62%; genistein treated, 3.11±0.79%; Fig. 6D). Given that neither AVE nor VE markers are affected by any of these inhibitors (see Fig. S6 in the supplementary material), these results indicate that proliferation is required for AVE migration and strongly suggest that it is proliferation in the epiblast that is required for these cell movements.

AVE cells have been shown to move away from a repulsive Wnt signal and towards a Dkk1 attractant signal (Kimura-Yoshida et al., 2005). A possible explanation as to how epiblast proliferation facilitates AVE migration is by moving the prospective AVE cells (DVE) away from the inhibitory Wnt signals. To test this hypothesis, embryos were cultured in the presence of nocodazole to inhibit proliferation and of Dkk1 as an AVE attractant. We found that only 3% (n=1/31) of nocodazole-treated embryos showed a migrated AVE and 35% (n=11/31) showed partial AVE movements, whereas in those embryos cultured in the presence of nocodazole and Dkk1 33% (n=11/33) showed a migrated AVE and 55% (n=18/33) displayed partial migration (Fig. 7). Interestingly, in the majority of these latter embryos migration seemed to have occurred unilaterally, suggesting that the cues for the direction of AVE movements had already been provided at the starting point of our cultures. These results indicate that the AVE migration defects caused by inhibiting proliferation can be rescued by Dkk1, and support the argument that the proliferation in the epiblast has a facilitative role for the AVE movements by moving the DVE away from signals that are inhibitory to its migration.

Little is known about how the growth of the embryo and patterning are integrated during early mammalian development. We have shown that epiblast proliferation is required to facilitate the initiation of AVE migration. This provides a mechanism to ensure that during post-implantation development, the rate of embryonic growth is tightly associated with the development of axial patterning.

Active migration (Srinivas et al., 2004) and passive displacement due to differential proliferation within the VE (Yamamoto et al., 2004) have both been proposed as forces that drive AVE movements. We have carried out a detailed analysis of the distribution of proliferation within the VE at the different stages of AVE migration and found that there is no evidence for differential cell division at any stage of this process between the anterior and posterior VE. A number of other studies support our findings. First, the DVE cells reach the embryonic/extra-embryonic boundary in as little as 4-5 hours, leaving cells behind at the distal tip of the embryo (Srinivas et al., 2004). Second, the average cell cycle length in the VE at the time of the DVE movements is estimated to be over 16 hours (Snow, 1977). Third, detailed analyses of clonal populations in the VE prior to gastrulation fail to outline major differences in proliferation rates between the four subregions of the distal VE (Perea-Gomez et al., 2001). In contrast to these findings, a previous study reported an absence of BrdU incorporation in the AVE, suggesting differential proliferation between the anterior and posterior VE (Yamamoto et al., 2004). It is possible that the difference between this and our study could be explained by a lower sensitivity of the protocol used to detect BrdU or variations in the genetic strains used.

The hypothesis that cell migration is the main motive force driving AVE movements is supported by the lack of any difference in proliferation rates in the VE, as well as by a number of other recent observations. Time-lapse imaging showing the dynamics of AVE movements (Srinivas et al., 2004), studies identifying attractants for these movements (Kimura-Yoshida et al., 2005) and the finding that embryos carrying mutations in genes that are essential for cell migration exhibit defective AVE movements (Migeotte et al., 2010; Rakeman and Anderson, 2006) all support this hypothesis.

Analysis of the distribution of proliferating cells at 5.5 dpc revealed that the right side of the embryo is proliferating at a significantly higher rate than the left side at this stage. Between 5.5 and 6.5 dpc there is a change in the orientation of the axis of bilateral symmetry of the embryo, an event that is thought to be required for the initiation of gastrulation (Mesnard et al., 2004; Perea-Gomez et al., 2004). Given that this change in axis of symmetry does not appear to involve cell movements (Perea-Gomez et al., 2004), it is tempting to speculate that it might be driven by differential proliferation between the left and right sides of the embryo.

During development, the growth of the embryo must be coordinated with its morphogenesis to ensure that patterning occurs in a correct and timely fashion. Here, we show that epiblast growth is directly linked to the initiation of AVE movements, an essential step for the correct establishment of the A-P axis.

Several lines of evidence suggest that proliferation in the epiblast is required for the initiation of AVE movements. First, as AVE movements are being initiated the growth imbalance between the epiblast and the VE grows from 2.2-fold to 11-fold and drops again once the A-P axis has been established. Second, after inhibition of Nodal signalling during the first phase of AVE migration, and in Cripto mutants in which the AVE forms but does not migrate, we see a decrease in the cell division rate in the epiblast but not in the VE. Third, genistein and hydroxyurea caused AVE migration defects and efficiently inhibited proliferation in the epiblast but failed to do so in the VE. These observations clearly indicate that proliferation in the epiblast is required for AVE migration. The specific loss of cell division that occurs in the epiblast after inhibition of Nodal signalling by SB-435142, as well as in Cripto mutants, clearly points to Nodal signalling as a major player in the control of embryonic growth. A critical target of Nodal signalling during this process is likely to be c-Myc, given the downregulation we observe in the epiblast expression of this gene after inhibition of Nodal signalling.

How does proliferation in the epiblast affect AVE movements? AVE cells have been shown to move away from a repulsive Wnt signal and towards an attractive Dkk1 signal (Kimura-Yoshida et al., 2005). We show that Dkk1 can rescue the AVE migration defects caused by inhibition of proliferation. This suggests that proliferation plays a facilitating role for AVE migration by moving the prospective AVE cells (DVE) away from signals that inhibit its migration.

The requirement for epiblast proliferation in the initiation of AVE migration is not the only control mechanism that ensures coordination of embryonic growth and A-P patterning. Indeed, the induction of the AVE requires a certain embryo size to ensure that the distal tip of the embryo is sufficiently distant from the inhibitory signals of the extra-embryonic ectoderm (Mesnard et al., 2006) and a critical cell number within the epiblast is required for the initiation of gastrulation (Power and Tam, 1993; Snow, 1977; Tam and Behringer, 1997). Therefore, it is becoming clear that during post-implantation development, a complex set of checkpoints exists to ensure that morphogenesis and growth go hand in hand.

In conclusion, we report that prior to, and during, AVE movements, there is no evidence for differential proliferation within the VE and it is therefore likely that cell migration is the driving force for AVE movements. We have identified a novel checkpoint control that links growth of the embryo with establishment of the A-P axis. This checkpoint ensures a coupling of the migration of the AVE to the Nodal-driven proliferation in the epiblast and therefore ensures that the patterning in both the epiblast and VE are coordinated.

We thank Michael Shen for the Cripto^+/− mice; Yi-Ping Li, Daniel Constam, William Shawlot, José Belo and Christof Niehrs for kindly providing plasmids; and Francis Vella for helpful discussion. T.A.R. holds a Lister Institute of Preventative Medicine Fellowship and this work was also supported by the Medical Research Council. Deposited in PMC for release after 6 months.