In recent years, the cascades of inductive interactions leading to neuralization and/or mesodermalization of naive embryonic tissues have been studied in great detail (Niehrs, 2004; Heasman, 2006). However, the problem of precise positioning and proportionality of neural and mesodermal rudiments is far from a solution. According to a widely accepted point of view, such positioning is fully provided by the balance between opposing concentration gradients of dorsal- inducing substances and the ventral factors, such as secreted proteins of the BMP superfamily (Reversade et al., 2005; Khokha et al., 2005). From this perspective, neural and mesodermal patterns should be firmly established already before the start of gastrulation (and, moreover, of neurulation), such that the role of morphogenetic movements is limited to placing predetermined rudiments in their proper locations. There are, however, several sets of arguments suggesting that this point of view is an oversimplification of a real situation. First, invagination movements, especially those of lateromedial cell convergence, are far from being precise when observed on the individual cell level; rather, they exhibit substantial stochasticity (Keller and Danilchik, 1988). Under these conditions, if cell fates were strictly determined before the start of morphogenetic movements, the resulting patterns would be highly variable, which is not the case. Next, Zaraisky (1991) showed that after extirpation of a substantial portion of non-axial tissues from X. laevis, embryos the final volumes of notochords, somites, and neural tissues were diminished in proportion to the reduced embryo volume, even if extirpations were made well after the start of gastrulation movements. The latest extirpation stages compatible with the proportional volume reduction were early gastrula for the notochord, midgastrula for somite tissue, and early-mid neurula for neural tissue. These results indicate that even up to the neurulation period, the borders between the axial and non-axial tissues as well as between different axial organs are not strictly determined and, therefore, gastrulation and neurulation movements can somehow participate in their precise positioning. Although this positioning can, in principle, be interpreted on the basis of BMP gradients without any involvement of morphogenetic movements (Ben-zvi et al., 2008), the corresponding models are essentially non-robust and applicable only to the interventions made at earlier developmental stages. Meanwhile, more robust pattern-generating models could be constructed by focusing on the tissue geometry created by morphogenetic movements (Beloussov and Grabovsky, 2006).
In our previous experiments (Kornikova et al., 2009), we arrested gastrulation movements by relaxing mechanical tensions in the suprablastoporal area (SBA) and traced the locations of tissue-specific genes expression sites in arrested embryos. These locations turned out to be rather variable and discontinuous and were far from coinciding with the maps of presumptive anlagen for pre-gastrula-stage embryos. These results argue for a direct participation of gastrulation movements in determining the precise patterns of axial organs. In addition, we noticed that neural rudiments and somite series were arranged as parallel arches along the extended lateral lips of the opened and curved blastopores, with the neural rudiments located along the inner (concave) surface and the somite series lining the outer (convex) side of the lips (Kornikova et al., 2009; see Fig. 7d). This neuro-mesodermal arrangement is the reverse of that predicted by anlagen maps, and led us to suggest that the mutual arrangement of axial rudiments is not strictly determined prior to gastrulation and can be influenced by experimental modulations of embryo geometry. The aim of this study was to test this idea in a more direct way. We prepared double explants of SBA ectoderm (later on defined as explants), artificially bent them either parallel or perpendicular to their antero-posterior (AP) axes, and compared the mutual arrangements of neural and mesodermal rudiments (identified by tissue-specific gene expression and conventional histology) in deformed explants with that of intact (non-deformed) tissues. We found that in spite of certain variability, the neuro-mesodermal patterns of intact explants obeyed a distinct axi-symmetric AP arrangement, with the neural rudiments shifted anteriorly. On the contrary, in the deformed explants that preserved the initial bending, the axi-symmetric patterns were completely disturbed: the neural tissue showed a definite bias towards the concave surface, while the mesodermal tissue tended toward the convex side, encircling the neural rudiment in a horseshoe-shaped fashion. In addition, some of explants, both intact and bent, took on a spherical shape. In these, the arrangement of the neural and mesodermal tissue was highly irregular and variable. By tracing cell movements and early morphogenesis in the artificially bent explants, we demonstrate that the imposed deformations trigger active extension of the explants' convex surface and active contraction/folding of the concave surface. We speculate that these movements play an important role in establishing a mutual neuro-mesodermal positioning in both SBA explants and during normal development.
In Vivo Traced Development of Artificially Deformed Explants
By time-lapse imaging, we traced the changes in the length of the ectodermal lining of the explants' opposite sides before and within a 3-hr period after bending (Figs. 1, 2; see Supp. Movie S1, which is available online). We found that in addition to a trivial extension of the convex lining and shrinkage of the concave lining caused by the bending itself, similar changes continued well after the explants were immobilized in their bent positions. Within the next 3 hr, the normalized length differences in the explants' opposite surfaces reached 40–70% (compared with only 12–33% differences immediately after bending) (Figs. 1A, 2A–C). In the intact explants, the corresponding differences were non-significant and did not exceed 10% (Figs. 1B, 2D,F). From this we conclude that the artificial bending serves as a trigger of active convex extension and concave contraction. This idea was confirmed by histological examinations of the early explants (see Histology of Recently Bent Explants and Measurements of Apical Indexes section) and by tracing fluorescently labeled samples.
In these latter samples, an initially compact label, placed on a convex part of the explant, split into several small patches, separated by large areas of non-labeled cells within 10–15 h (Fig. 3, compare A and B, C and D). The distances between the most widely separated patches were 3–5 times greater than the diameter of the initial label. This result points to extensive lateral cell spreading, associated with elongation of the convex side of the explant. A label placed on the concave surface did not spread at all but was instead concentrated around an invaginated zone (Fig. 3. compare E and F). In none of the several dozens experiments did we observe any traces of the vertical migration of the labeled cells from one side of explant to the other.
Histology of Recently Bent Explants and Measurements of Apical Indexes
Five minutes after the bending was complete, one or two shallow folds had formed on the explant's concave surface (Fig. 4A). Thirty minutes later, one of these folds dominated and was transformed into a narrow ingression, with a single bottle-shaped cell onto a tip (Fig. 4B, B2, bc). The opposite surface (Fig. 4B1) did not significantly differ from control intact explants. Three hours later, the ingression in the explants became curved, a number of bottle-shaped tip cells increasing (Fig. 4C, C2). Below the convex surface, the first stages of somite formation could be traced (Fig. 4C1). In several bent and intact explants, a miniature spherical blastocoel had formed (Fig. 4C).
As seen in Figure 5, a difference in the average apical index (AI) of the convex and concave surface cells of the bent explants became apparent by 5 min after bending and continually increased thereafter, with a more than 5-fold difference observed after 30 min (P < 0.001). The cells on the concave side adopted bottle-like shapes with narrow apices, while cells on the convex side approached cuboid shapes. In contrast, the AIs of the intact explants did not differ significantly from those of the intact embryonic ectoderm tissues used for explants preparation. Taken together, all of these data indicate that explants' bending triggers a series of active and rapid responses directed towards the reinforcement of increased curvature.
Mechanical Stresses in Bent Explants
In the following experiments, we used a “mechanical jumps” test that can provide sufficient evidence for the existence of residual mechanical stresses in embryonic tissues (Hutson et al., 2003). The first mechanical jump is a rapid backwards straightening of a released explant kept in a bent position for no more than a few minutes. This reaction means that during the bending procedure explants were substantially stressed by the bending force. The internal mechanical balance was restored, due to cells rearrangements, a few dozen minutes later when the explant liberated from external constraints did not exhibit any rapid deformations. Meanwhile, restoration of such a balance in no way means that the stresses within the explants' tissues become negligibly small; rather, incisions made on the convex side of 3-hr explants induced another mechanical jump, namely, a rapid widening of the wound gap (Fig. 6A), indicating tension on the explant's surface. In 20-hr explants, the stress patterns became more complicated. On the one hand, the ectodermal edges of an incision (Fig. 6B, pointers) separated to a much greater distance than in the early explants, indicating rising surface tension. On the other hand, the internal cell mass did not follow the ectodermal opening as it did earlier, but instead closed the wound gap, demonstrating the release of a pressure stress (Fig. 6B, double arrows). In addition, within a few dozen seconds the curvature increased in the surface opposite to incision. We conclude that at this time the overall stretching of the explant's surface is combined with the expanding pressure of internal cell mass.
Arrangement of Neural and Mesodermal Rudiments, as Revealed by In Situ Hybridization of 20–24-Hour Intact and Bent Explants
Intact explants could be divided into two categories. First, despite some shape variability, most (14 out of 18) preserved a definite antero-posterior (AP) polarity: neural rudiments were located close to the anterior pole while mesodermal tissues and the notochord occupied the posterior zone (Fig. 7). The average asymmetry index (see Experimental Procedures section) for neural tissue was close to zero (A = 4.4 ± 3.5; n = 19; t = 5.4), indicating the preservation of axial symmetry. The remaining four explants lost AP polarity and took spheroid shapes. In these, the arrangement of neural and mesodermal rudiments was quite variable and irregular (Fig. 8A–C).
Among the bent explants (Table 1), in spite of the imposed deformation, 8 out of 49 resumed a symmetrical, elongated shape. These showed distinct AP polarity and did not differ significantly from the intact samples. Sixteen explants took spheroid shapes and, similar to the intact ones having the same shape, showed irregular neuro/mesodermal arrangements (Fig. 8D–F). The remaining 25 explants maintained the imposed bending. In these, AP-polarity was lost and, irrespective of the bending direction, the arrangement of neural and mesodermal rudiments was highly asymmetric and curvature-dependent (Fig. 9A–J). To evaluate the results of effective bending, we used three different quantitative criteria. First, by signs criteria (Van der Waerden, 1957) the fraction of samples with Sox3 expression sites fully located within the concave half of the explant exceeded the fraction with convex-biased Sox3 localization with 99% significance. Next, an overwhelming majority of samples (22 out of 25; significance by signs criteria 99.5%) possessed horseshoe-shaped mesodermal rudiments that embraced neural tissue from the convex side. This arrangement could also be traced in some spherical explants (Fig. 7F). Finally, the average index of asymmetry for neural rudiments was highly positive (+ 68 ± 44) and the difference from the intact explant index (4.37 ± 3.5) was highly significant (P < 0.0001). Otx2 expression sites were also located in non-convex areas, either flattened (Fig. 10A) or concave (Fig. 10B–E, arrows). At the same time, they tended to shift towards one of the explant poles (Fig. 10C–E), even if the latter did not coincide with the anterior pole of the explant. In the extensively elongated and straight explants (Fig. 10E, F), this tendency was quite pronounced, but even in these cases it was clear that Otx2 expression sites coincided with transversal grooves, i.e., they were located in concave regions (Fig. 10F).
Table 1. Summary of the Operation Results
Direction of bending
Numbers of explants from different experimental series
Kept straight, elongated in AP direction
Adopting spherical shapes
Preserving imposed bending Among them:
Neural tissue located within the “concave half” only
Neural tissue located within the “convex half” only
Mesodermal tissue horseshoe-shaped, encircling neural tissue from the convex side
Average asymmetry index for the neural tissue
+ 68 ± 44 (n = 14; t = 5,7)
Sites of Neural and Mesodermal Differentiations in SBA Tissues Depend Upon Explant Geometry
The main aim of this work was to explore whether a canonical mutual arrangement of neural and mesodermal rudiments in Xenopus embryos could be modified at the early-midgastrula stage by geometry-dependent morphogenetic processes, triggered by artificial deformation of suprablastoporal ectoderm. As an adequate control, we used non-deformed (unconstrained) explants from the same area. We observed that in most of the controls (with the exception of those adopting a spherical shape), the neuro-mesodermal arrangement obeyed a distinct AP polarity (neural rudiments located to the anterior of the mesodermal tissues) and was axi-symmetric. This arrangement roughly corresponded to the anlagen maps for a given area, though not in minor details. In addition, this arrangement did not significantly differ from that of standard Keller's explants, which, in addition to our SBA explants, also contained a marginal zone of the dorsal blastoporal lip (Wilson and Keller, 1991).
The neuro-mesodermal patterns observed in the artificially bent explants that preserved the imposed bending were drastically different. Independent of the bending direction (oriented either parallel or perpendicular to the AP axis), the initial AP polarity and the axial symmetry of both the neural and mesodermal rudiments were lost in all these cases. Instead, in spite of the distinct cellular identities of the convex and concave halves of the double SBA explants, we observed a statistically significant bias of neural rudiments towards the concave surface and mesodermal tissues towards the convex. We associate these abnormal locations with specific patterns of cell rearrangements triggered in the opposite sides of explants by imposed bending. That is, as shown in Figures 1 and 2, by several minutes after bending the convex part of the explant started to extend while the concave side began to contract. On the convex surface, this shift was accompanied by a reduction in the apical indexes of the ectodermal cells (Fig. 5) and extensive lateral spreading of the deep cells (Fig. 3). In turn, cells on the concave side acquired bottle-like shapes with high apical indexes (Fig. 5) and migrated inside the explant, forming a deep fold (Fig. 4B, C). Both processes, which considerably reinforce the imposed bending, are clearly active since they proceed for a prolonged time after the end of the artificial deformation of a sample. This may be qualified as active reinforcement of externally imposed curvature, suggested by the “stress hyper-restoration” model (Beloussov, 2008). Recent data from our group (in preparation) indicate that not only SBA explants but also other kinds of embryonic tissues are capable of similar reinforcement of an imposed curvature. However, only in the case of SBA explants (consisting of tissue competent to form axial organs) do these kinematic responses affect subsequent cell differentiation.
From this point of view, the extensive irregularity of the neuro-mesodermal arrangement observed in spherical explants may be regarded as a consequence of their uniform curvature, providing no geometrical cues for regular positioning of the two differentiated cell types. This irregularity again demonstrates once more that all of the SBA cells at the early-midgastrula stage are still in a state of instability permitting persuasion towards either of the alternative differentiation pathways.
What are the alternative ways for interpreting our results? First, one may assume that the cell convergence and folding observed on the concave side of the bent explants is a consequence, rather than a prerequisite, of neural differentiation, because the normal neurulation is accompanied by folding. However, this suggestion cannot be accepted for the following reasons: (1) the active folding on the concave side starts very early (not later than 30 min after bending) while the folding associated with normal neurulation takes place several hours later (at the early neurula stage of intact embryos); (2) the bulk of the neural tissue developed in the intact SBA explants (e.g., Fig. 7H) did not show any signs of folding; (3) even if we accept, contrary to (1) and (2), that the folding is a consequence, rather than a requirement of neural differentiation, the fact that the neural tissue developed specifically on the concave side remains unexplained.
Next, one might suggest that the observed locations of neural and mesodermal rudiments in the bent explants are caused by secondary shifts of these tissues, which were committed to their specific differentiation pathways while still occupying their “canonical” polar positions. This assumption is also somewhat unrealistic because it implies extensive mixing of cells between the two sides of the explant (i.e., committed neural cells on the convex surface would have to migrate towards the concave side, while mesodermal cells traveled in the opposite direction). However, using fluorescent labeling, we did not observe any such movements, even in a 20-hr observation period, while the first signs of geometry-dependent differentiation (onset of somite formation on the convex side, see Fig. 4C) could be traced no later than 4 hr after bending. Therefore, we conclude that the cells initially located on either the concave or the convex side of the explant were recruited to the neural or mesodermal differentiation pathways, respectively.
It is worth mentioning in this respect that the rapid forcible bending used in our experiments can in no way be replicated by the slow spontaneous curving often observed in the posterior parts of intact explants (Fig. 7D). This type of curving, also ubiquitous for the standard Keller's explants (see http://www/apple.com/quicktime/) is probably a consequence, rather than the cause, of the slight axial asymmetry in the actively extending mesodermal (notochordal) rudiments. Due to its slowness, this event can hardly play any role in reprogramming cell fates.
Interestingly, several reports from the past 75 years contain photographic evidence (Spemann, 1936, fig. 153; Saxen and Toivonen, 1963, figs. XXVI, XXVIII; Saint-Jeannet et al., 1994, fig. 1B; Grunz et al., 1995, fig. 3) of the preferential localization of neural tissues in the curved (concave) areas of different embryonic preparations (e.g., exogastrulae, induced sandwiches, Keller's explants). However, as the time and rate of the curving in these cases was not documented, it is impossible to determine whether the changes in geometry were primary or mere consequences of preceding differentiation.
Fundamental Similarities in the Mechano-Geometry of Bent Explants and Normal Neurula-Stage Embryos
One of the first processes involved in neurulation is the longitudinal extension and transversal shrinkage of neuroectoderm (Jacobson and Gordon, 1976). The required longitudinal extensive force (Fig. 11B, 1) is produced by lateromedial convergence of the axial mesoderm cells, associated with notochord formation (Keller and Danilchik, 1988). The longitudinal pressure stress generated by this force can be visualized by the rapid substantial overlapping of the edges of transversely dissected gastrocoel roof in the early-mid-neurula Xenopus embryos (Beloussov et al., 1994; see fig. 2b). By mechanical laws, this very force not only longitudinally stretches the adjacent parts of the lateral embryo wall (Fig. 11B, 2), but should also produce a transversal shrinkage of neuroectoderm (so-called Poisson's deformation) (Fig. 11B, 3). Further creasing of neuroectoderm can be produced by ventrodorsal pressure, which has been proposed to be exerted by the lateral ectoderm (Colas and Schoenwolf, 2001). In any case, before the start of the active bending movements, the neural plate is likely under a degree of transversal compression, similar to the concave parts of our deformed explants. Therefore, a succession “passive creasing-active contraction” should take place during normal neurulation as well, although in that case oriented exactly in the transversal plane (Fig. 11, compare A and B). At the next stage, the active bending and rolling of the neural plate stretches the entire lateral embryo surface in the dorsoventral direction (Fig. 11B, 5). This stretching should in turn trigger the active ventrodorsal cell movements (Fig. 11B, 6) similar to those taking place on the convex side of the bent SBA explants. The mechanodependence of the ventrodorsal cell movements has been demonstrated by their suppression in mechanically relaxed embryos (Beloussov et al., 1990).
In summary, one can see that the mechano-geometry of normal neurulation, as seen in the transverse plane, is much the same as in the artificially bent double explants. This association is supported by the similarity of the neuro/mesodermal arrangement in the bent explants and intact embryos (Fig. 11, compare C and D).
At the same time, the presence of extensive longitudinal stresses in intact embryos (and to some extent in the intact explants) adds several features lacking in the bent explants. In addition to the aforementioned longitudinal stretching (Fig. 11B, 2), this additional stress is apparent in the longitudinal compression of the anteriormost (forebrain) part of the neural plate (Fig. 11B, 7), which, by our suggestion, triggers the subsequent active longitudinal contraction (Fig. 11B, 8). Obviously, in the bent explants these axial activities are broken and antero-posterior polarity is therefore inhibited.
In spite of some remarkable claims for the importance of morphogenetic movements in the specific regional differentiation of the induced rudiments (Yamada, 1994), until recently the general consensus was that morphogenetic movements and associated mechanical stresses are mere consequences of previously established patterns of genes expression, and are thus unable to affect the latter. Impressive results on mechanosensitivity of twist expression in Drosophila embryos (Farge, 2003) demonstrated, however, that this view was one-sided, and that the reverse situation can also occur and be of essential importance. As neural and mesodermal rudiments were distinguished by the criteria of tissue-specific gene expression in the present study, we regard our data as more evidence for the mechanodependence of the genes involved in early development. Our data are also in line with some recent findings in stem cells studies (McBeath et al., 2004; Engler et al., 2006) emphasizing the crucial role of cell shape and the mechanical properties of substrate for determining the direction of cell differentiation.
Xenopus laevis (Daudin) embryos obtained from hormonally stimulated adults and incubated at room temperature were used in this study.
Before operations, embryos were dejellied with a 2.5% cystein solution and liberated from vitelline membranes with forceps. During and after operations, embryos were incubated in small Petri dishes onto 2% agarose substrate, filled by Marc's modified Ringer's (MMR) solution (100 mM NaCl, 2 mM KCl, 2 mM Ca Cl2, 1 mM Mg Cl2, 5 mM HEPES, pH 7.4). At the early to midgastrula stages (10½–11 according to Niewkoop and Faber, 1956), rectangular (about 500- × 500-μm) pieces of a suprablastoporal ectoderm were fused by their inner surfaces in pairs with the same AP polarity. The fused pieces were similar to what is usually defined as Keller's explants (e.g., Wilson and Keller, 1991) but did not include bottle-shaped cells in the area adjacent to the dorsal blastoporal lip. After about 30 min, the fused explants were inserted sideways in slightly curved beds dug by glass sticks into the agarose substrate. To bend the explants parallel to their AP axes, they were mounted in the beds with AP axes oriented vertically, while to bend them perpendicularly to the AP axis (that is, transversely), they were mounted with AP axes oriented horizontally (Fig. 12). Most important for successful bending was to use short beds: under these conditions explants curved when AP axes were oriented horizontally. Explants were incubated for different time periods up to 20–22 hr and then fixed either for in situ hybridization or conventional histology.
Time-lapse filming was performed with the use of digital camera DCM130, 1.3M pixels. Frames interval was 1 min.
To trace cell movements in double explants, one of the fused tissue pieces was taken from an embryo in which one of the dorsal blastomeres had been injected with 10 ng of fluorescein-dextran (FD) (40,000 MW, Molecular Probes, Eugene, OR) at the 16-cell stage. At varying time points after explant fusion, fluorescence was monitored under UV excitation with the use of an Olympus SZX9 epifluorescent microscope with fluorescent block U-RFL-T.
Semithin Epon sections of 4-hr incubation time double explants fixed in 2.5% glutaraldehyde and 7–10-μm paraffin sections of 20–22-hr incubation time double explants fixed in Bowen fluid were prepared by routine techniques (for details, see Beloussov et al., 2000).
Testing Residual Tensions
To evaluate residual tensions in embryonic tissues, we made local incisions, fixed the samples in Bowen fluid for a few seconds after operating, then traced the width of the operational gaps, assuming that the residual tensions are roughly proportional to the width of the gap (see Hutson et al., 2003).
In Situ Hybridization
In situ hybridization was performed with standard technique (Harland, 1991), and was used to reveal expression patterns of the muscular actin gene, the pan-neural gene Sox3 and the anterior neural gene Otx2. Antisense probe against Xenopus cardiac actin was prepared from pAC100 (Mohun et al., 1984). PCR product from the plasmid, amplified with Sp6 and T7oligos, was transcribed with SP6-RNA polymerase. Antisense probe against Xenopus Sox3 was prepared from the plasmid kindly provided by Dr. Grainger (Zygar et al., 1998). The plasmid was cut with SmaI and transcribed with T7-RNA polymerase. Antisense probe against Xenopus Otx2 was prepared from pXOT30.1 (Lamb et al., 1993). PCR product from the plasmid, amplified with T3 and T7oligos, was transcribed with T7-RNA polymerase.
For double in situ hybridization, the hybridization procedure remained generally the same. Digoxigenin-UTP-labeled (Sox3) and fluorescein-UTP-labeled (cardiac actin) RNA probes were used. Embryos were incubated with anti-DIG antibody (Roche, Nutley, NJ) diluted at 1:2,000 in antibody incubation solution (MAB + 3% powder milk + 20% heat-inactivated lamb serum) at 4°C overnight. The first antibody was detected with Purple AP Substrate (Roche). Following heat-inactivation of the alkaline phosphatase (10 min at 65°C in MAB with 10 mM EDTA) and two 10-min washes in MAB, the embryos were blocked in 3% powder milk in MAB for 1 hr at room temperature. Embryos were then incubated with anti-fluorescein antibody (Roche) diluted at 1:1,500 in antibody incubation solution (MAB + 3% powder milk + 20% heat-inactivated lamb serum) at 4°C overnight. Following extensive washes in MAB, the second antibody was detected with Fast Red tablets AP substrate (Roche).
Morphometric measurements were made with the use of the ImageJ program. Apical indexes (AI) were evaluated according to Lee and Harland (2007) as the ratios of maximal radial cell length to the same cell apical diameter. To estimate the asymmetry of gene expression patterns, images of in situ hybridized samples were divided by midlines equidistant from the opposite explant surfaces as shown in Figure 13 and the areas of Sox3 expression located on the opposite sides of the midline (S1 and S2) were measured in relative units. The asymmetry index (A) for each individual sample was calculated (in percent) as A = (S1 − S2) / (S1 + S2) × 100 (100 > A > 0). Only samples of a uniform thickness were taken for measurement. While S1 and S2 could be set only arbitrarily in the intact axi-symmetric samples, in the bent samples we defined S1 as the area situated between the midline and the concave side of the explant. In these cases, A-s could be either positive or negative, with positive values corresponding to an excess in the concave side areas over convex sides areas.
Statistical evaluations were made with the use of Statistica 6.0 program, section Basic Statistics. The significance of experimental results was evaluated by a “signs criteria” (Van der Waerden, 1957).
The authors thank Dr. L.A. Berekelya for her skilled consultations.