Cellular Morphogenesis of Ascidians II: Notochord Morphogenesis
(Convergent Extension)


Cellular morphogenesis is a fundamentally mechanical process in which many individual cells move and change shape, exerting forces upon and experiencing forces from their immediate neighbors, and in so doing they drive tissue-level deformations. None of an embryo's cells know what the group as a whole is supposed to do or make. So how do they manage collectively to reorganize and deform themselves in a stereotyped way, producing just the right outcome every time?

In recent years, cell biologists have begun to uncover and characterize the machinery that individual cells use to generate local contractile or protrusive forces, to polarize, to establish and transmit forces through local adhesive contacts with one another or with extracellular substrata, and so on. This core machinery turns out to be strikingly conserved within the animal kingdom.

Given this conservation, it seems likely that same core machinery that propels isolated cells in vitro also drives the movements of embryonic cells during morphogenesis. The fundamental question then becomes: How does conserved cytoskeletal machinery operating within each of many cells to generate purely local forces, cause a population of cells, whose members are interlinked by adhesive machinery, to reorganize and deform in a characteristic pattern? How do differently fated embryonic cells regulate the local force-generating and adhesive machinery differently so that they collectively produce a stereotyped global outcome? How has the evolutionary process tuned this machinery to produce morphogenetic variation across phylogeny?

To address these questions, requires comprehending how local force-generating processes operating at cellular or subcellular levels yield tissue-level consequences. Physical law dictates that, at every moment, forces arising locally anywhere within an embryo resolve themselves simultaneously and globally into instantaneous patterns of cell movement and deformation. It is often difficult to intuit the outcome of this resolution. At the CCD, we combine detailed computer simulation models with modern imaging tools, micromanipulation, and molecular perturbation, to approach a mechanistic understanding of how some specific morphogenetic machines actually work.

Our recent work has focused on understanding the morphogenetic machine that drives the formation and extension of the notochord, and in related work, of endoderm invagination, and for both studies we avail ourselves of the many advantages of ascidian embryos. The ascidian notochord forms from a simple monolayer sheet of forty cells that reorganize themselves in about 6 hours without cell divisions to form an extended rod of coin-shaped cells (Munro and Odell 2002a&b).

Schematic view of different stages of notochord formation in an ascidian embryo.
Left panel: early gastrulation; Middle panel: late gastrulation; Right panel: tailbud stage.
Presumptive notochord cells are shown in red, endoderm yellow, muscle orange, epidermis gray, and neural plate blue.

This transformation involves two simultaneously occurring processes: invagination of the notochord plate forms a cylindrical rod; and intercalation of cells across the plate and about the circumference of the rod causes it to become longer and thinner.

Schematic view of the two morphogenetic prcesses that simultaneously transform the notochord plate into a cylindircal rod.

During the 6 hours in which these events occur (and not before or after), individual notochord cells crawl actively across the basolateral surfaces of their notochord neighbors. This crawling appears to be polarized with respect to both the axis of notochord extension and the apico-basal axis of individual notochord cells: cells appear to crawl more actively along the axis of intercalation (perpendicular to the axis of notochord extension) and they appear to crawl more actively near the basal surface of the notochord. These observations suggest that active local motility, expressed identically by all notochord cells and polarized with respect to the apico-basal and AP axes of the notochord plate, could generate the asymmetric global distribution of forces necessary to cause the invagination of the notochord plate on the one hand, and the convergent extension on the other.

Notochord cells converge and extend by crawling locally across the basolateral surfaces of their adjacent neighbors in
much the same way that an isolated fibroblast would crawl across an external adhesive substratum.


Modelling convergent extension of an epithelial sheet

Based on these observations, we hypothesized that the same conserved cytoskeletal and adhesive machinery that endows a single isolated cell with the ability to crawl across an isolated adhesive substratum could account, both qualitatively and quantitatively, for the ability of a sheet of the same cells to converge and extend or invaginate. To see whether and how this might be true, we built a computational framework to predict, from specific assumptions on the conserved contractile, protrusive and adhesive mechanics operating locally within each of many cells, what a sheet of such cells would do. The basic idea was build individual model cells from a collection of discrete elements- each representing a small piece of the cell cortex or internal cytoplasm, the nucleus, adhesion proteins, and so on- and to endow each of these discrete elements with local kinetic and mechanical properties designed to mimic those measured empirically for their real counterparts. We wanted to build in only the local behaviors and let the global consequences emerge. To model a sheet of cells, we simply arrange many model cells as neighbors and then allow them to adhere.

The top tryptich shows how local protrusive, contractile and adhesive machinery might operate within an individual cell
crawling across an isolated substratum. Below, a hypothetical view of how the same machinery might operate within a
monolayer sheet of embryonic cells during convergent extension.

Initially, we focused on three fundamental cytomechanical behaviors, thought to be necessary for the directed movements of single cells across isolated substrata (e.g. Alberts et al 2003, Bray 2000): 1) localized actin-dependent protrusive extension of the leading edge; 2) heterophilic or homophilic adhesion mediated by reversible binding between proteins located on adjacent cell surfaces and dynamically anchored to the underlying cortex; and 3) actomyosin-based contraction of the cortex. When we endowed model cells with these three local behaviors, tuned parameters governing contractile, adhesive and protrusive mechanics to match empirical measurements, and biased protrusive extension to a single "leading edge", we found that a single model cell would crawl in a realistic way across a rigid external adhesive boundary. Thus the textbook model of how individual cells crawl works mechanically.

Click on the panels above to see the QuickTime movies of a simulated cell crawling across a flat adhesive substratum.
Left: low resolution view. [334 KB] Right: High resolution view. [2.2 MB]

To our surprise, we found that a monolayer sheet of the same cells, identically tuned, with protrusive extension biased medially and laterally, would converge and extend! Moreover, the basic convergent extension behavior was remarkably robust with respect to quantitative variation of parameters governing protrusive, contractile and adhesive mechanics.

Click on each of the panels above to see the timelapse movie.
Left: The real thing
, filmed within an intact embryo [opens in new window 360 x 360 pixels, 3.7 MB];
Middle: Model simulation
showing how a sheet of model cells executes convergent extension [opens in new window 680 x 630 pixels, 1.4 MB];
Right: A close up view
of one three cell junction from the same simulation [opens in new window 680 x 560 pixels, 1.5 MB].


Some Future Directions

  • Testing the model. The model described above makes a number of very interesting predictions that we are within easy reach of testing. For example, it predicts qualitative patterns of cortical flow that we would expect to observe within the notochord plate if the assumptions we have made about the local cytomechanics that drive convergent extension are true. We plan to test these predictions in ascidian embryos by monitoring cortical flow in living embryos during convergent extension, either by labeling cells with extracellular fluorescent beads, or expressing photoactivatible markers within the notochord lineage. As another example, our model makes the (to us at least) counterintuitive prediction that the rate of convergent extension should increase with increasing adhesivity of notochord cells. We expected intuitively that the opposite would be true because we reasoned that notochord cells would somehow have to peel neighbors apart in order to move between them. We would like to test this prediction by varying extracellular calcium to vary (calcium-dependent) cell-cell adhesion within notochord explants and monitor its effects on the rate of convergent extension.
  • How does the global polarity of convergent extension emerge? So far, we have shown that a sheet of cells will converge and extend robustly if we bias local protrusive activity to medial and lateral edges to mimic what we observe in situ. But the $10,000,000 question is: how does this axis of polarity emerge normally? We have found that the ascidian notochord intrinsically manifests a robust tendency to converge and extend along some axis, but the orientation of that axis depends on interactions with surrounding tissues (Munro and Odell 2002b). None of the mechanics built into the model described above would compel a sheet of cells to do this. Thus it is a question of both experimental and theoretical interest: Whence this intrinsic tendency of cells to polarize? Is it intrinsic to individual cells or does it only emerge within a sheet of cells? How do interactions with neighboring tissues constrain or bias this tendency to produce a specific global axis of convergent extension? Recent studies from other labs suggest that a version of the planar cell polarity network first discovered in Drosophila embryos may be involved in generating polarized convergent extension, both in ascidians and in vertebrates (e.g. Wallingford et al 2002, Keys et al 2002). We are planning to incorporate regulatory network dynamics ala Ingeneue into our cytomechanical models to explore how local regulatory circuitry might couple with cytoskeletal and adhesive mechanics to govern polarized convergent extension.
  • Do the same local cytomechanics drive the convergent extension and invagination of the notochord plate? Thus far our modeling efforts have focused on understanding the cytomechanics of convergent extension. But as we described above, the ascidian notochord executes two morphogenetic movements simultaneously- invagination and convergent extension. It would be amazing if these two processes did not utilize some of the same machinery. Indeed, it's interesting to speculate that the very same local protrusive, contractile and adhesive mechanics that drive convergent extension when biased within the plane of the notochord plate, could also drive invagination of the plate if biased along the apico-basal axis. To explore this idea computationally will require going beyond the two dimensional models we have made thus far.
  • Cytomechanics of cell adhesion. Computational models often yield unlooked for insights into the nature of some process. Modeling convergent extension forced us to think hard about adhesion and how the adhesive forces that draw or hold cells together would act upon a deformable cortex. The insight which struck us is that any mechanism of adhesion which relies upon the transmission of forces between cells by elastic elements will, given the essentially viscous nature of the cell cortex on long time scales, induce cortical flows. Thus it becomes interesting: a) to detect and measure these flows experimentally in adhering cells; and b) to explore computationally and experimentally some of the potential consequences of these flows- for example how they might contribute to reorganizing the cortex or redistributing proteins within the plane of the membrane. We are beginning to explore some of these issues, and we are also planning to incorporate more detailed representations of adhesive mechanics into our models.

Page written by Ed Munro
Last updated March 20, 2008