Tensions divide

During development, proliferating cells are organized into compartments with boundaries across which cells fail to intermix. Compartment boundaries are often attributed to differential cell–cell adhesion between separate compartments. However, tension generated by actomyosin cables at boundaries can also function as a barrier that prevents cell mixing.

During embryonic development, cells are partitioned into distinct groups or 'compartments' separated by sharp, often invisible, boundaries (Fig. 1a). Cells fail to intermix across compartment boundaries, ensuring that their fates and/or positional information remain segregated as they proliferate and move. Because of their precise positioning, compartment boundaries often serve as signalling interfaces that function as sources for morphogens that pattern the surrounding tissue¹. Thus, the establishment and maintenance of compartment boundaries is of critical importance to embryonic pattern formation and tissue differentiation.

Figure 1: Mechanisms that establish compartment boundaries during development.

(a) Schematic diagram of anterior (A) and posterior (P) compartments in both the Drosophila wing imaginal disc and embryo. Dotted lines indicate compartment boundaries. In both the wing disc and the embryo, P-type identity is specified by the transcription factor Engrailed (En). En-expressing cells secrete Hedgehog (Hh) signal that activates the transcriptional regulator Cubitus interruptus (Ci) and A-type identity. (b) Three contrasting models of compartment formation. Differential adhesion: quantitative differences in cell–cell adhesion cause different cell/tissue types to have different tissue surface tensions (arrows along interface). By analogy to liquids, tissues with different surface tensions are immiscible. Differential contractility: different tissue surface tensions are not due to differential adhesion, but to different levels of cell contractility between cell types. Interfacial actomyosin cable: in contrast to differential adhesion and differential contraction, which result from the mechanical properties of the group, a boundary-specific actomyosin cable results from interactions between different cell types. Increased mechanical tension at the boundary restricts cell movement. Junctions could be required to transmit tension between interfaces along the compartment boundary.

Compartment formation is most often attributed to differences in cell affinity. This concept originated from work on amphibian embryos where it was observed that dissociated cells self-assemble with cells of a similar fate and sort into immiscible layers². These observations led to the hypothesis that cell sorting and compartment formation rely on differential adhesion, such that cells and tissues minimize their free energy and surface area by maximizing cell–cell adhesive interactions³. Thus, like liquids, tissues should exhibit surface tension that is proportional to the adhesive energy between constituent cells, and tissues with different surface tensions should fail to intermix (Fig. 1b). In support of this differential adhesion hypothesis (DAH), cell aggregates with different levels of cell–cell adhesion have different surface tensions and are immiscible⁴.

The DAH predicts that boundaries are a consequence of bulk adhesive differences between separate compartments. However, in vivo evidence showing that differential adhesion controls cell behaviour at compartment boundaries is lacking. Genetic screens have so far failed to identify adhesion molecules that distinguish cells in different compartments or account for behaviours at the compartment boundary. In addition, although assays have been developed to measure cell–cell adhesion in dissociated cells⁵, it has not yet been possible to quantify cell–cell adhesion in vivo. Thus, other mechanisms could contribute to compartment formation.

Cell sorting and compartment formation may also be influenced by actomyosin contractility. Adjacent cell types with different levels of contractility could create differential surface tension and drive cell sorting (Fig. 1b)^5,6. Alternatively, rather than simply resulting from bulk mechanical differences between cell types, boundaries themselves could have specialized mechanical properties that actively restrict cell movement. For example, filamentous actin (F-actin) and non-muscle Myosin-II (MyoII) are enriched at the dorsal–ventral (D–V) compartment boundary in the Drosophila wing disc⁷. Two new studies, one on page 60 of this issue, support the possibility that this accumulation increases tension along compartment boundaries, and forms a physical barrier that prevents dividing cells from entering the adjacent compartment^8,9.

The Drosophila wing disc is subdivided into anterior–posterior (A–P), as well as D–V, compartments (Fig. 1a)^1,10. Although upstream signalling molecules and transcriptional regulators that specify A and P identity have been identified (Fig. 1a)¹¹, the effector molecules that establish this compartment boundary were unknown. Dahmann and and colleagues have found that F-actin and MyoII are enriched on the border between A-type and P-type cells⁸ and demonstrated that this localization reflects increased tension along the compartment boundary, using two approaches. First, quantification of angles between linked interfaces along the A–P boundary showed that they were higher, or straighter, than linked interfaces away from the boundary. Second, the authors ablated cell interfaces and measured vertex displacements to estimate relative tensions at interfaces. This analysis revealed that interfaces along the A–P boundary exhibit ∼2.5 times higher tension than interfaces within a given cell type.

Compartment boundaries are also observed during early stages of Drosophila embryonic development, when the main body segments of Drosophila are divided into lineage-restricted A and P compartments (Fig. 1a). Sanson and colleagues investigated how this boundary is maintained by screening for chromosomal deletions that disrupt boundary formation⁹. They identify MyoII as a critical effector of boundary formation, similar to wing disc boundaries, and demonstrate that it is enriched along A–P boundaries, forming cable-like structures. To determine how actomyosin contractility might restrict cell mixing, the authors performed live imaging of this proliferating tissue. They found that cells push into the neighbouring compartment when they divide. Surprisingly, the actomyosin cable is maintained during cell division, even when a contractile ring is assembled, and seems to resist the movement of newly generated cells away from their compartment of origin. This suggested that actomyosin tension forms a barrier that maintains the compartment boundary when challenged by cell division.

Analysing whether MyoII accumulation is responsible for restricting cell movement is complicated by its pleiotropic role. MyoII is required for cell–cell adhesion¹², making it difficult to functionally separate cell–cell adhesion and contractility mediated effects. In addition, globally inhibiting MyoII could disrupt cell division and thus block the process that initiates cell mixing. To circumvent this problem, the authors used chromophore assisted light inactivation (CALI) to locally and specifically inhibit MyoII activity along the compartment boundary. In CALI, intense irradiation of a fluorophore, such as green fluorescent protein (GFP), generates reactive oxygen species that specifically inactivate a tagged target protein that is within 50 Å. The authors showed that irradiation of MyoII–GFP, but not a GFP-tagged actin-associated protein, precisely inhibits actomyosin contractility within a cell. CALI of MyoII at the compartment boundary allowed dividing cells to invade the neighbouring compartment, resulting in an irregular compartment boundary. This is the best evidence so far that actomyosin contractility has a role in cell sorting that is independent from adhesion, as interfaces that contact 'like' cells, which could mediate sorting based on adhesion, are not targeted. Thus, the actomyosin cable seems to serve as a 'fence' that restricts cell movement between compartments during proliferation (Fig. 1b).

As mentioned earlier, previous models for cell sorting predict that boundaries form as a consequence of different cell types having different adhesive or contractile properties (Fig. 1b). However, several lines of evidence suggest that A- and P-type cells actually have similar mechanical properties. First, in the wing disc, interfacial tensions within the A and P compartments are identical⁸. Because contractility and adhesion both contribute to interfacial tension, it is possible that differences in contractility are compensated by differential adhesion and vice versa. However, similar levels of cortical MyoII and E-cadherin are observed in A and P compartments, suggesting that this is not the case^8,9. Second, the actomyosin cable appears on both sides of the A–P boundary in segments, suggesting that both cell types assemble this specialized contractile structure⁹. Third, boundary formation can be recapitulated using an energy model, in which an increase in the interfacial tension between cell types is sufficient to drive cell sorting between mechanically similar cells⁸. Thus, these compartments do not seem to result from mechanical differences between cell types, but possibly result from the mechanical properties of the boundary itself.

The requirement of actomyosin contractility does not preclude an important role for cell–cell adhesion. Adherens junctions are probably needed to link A–P interfaces such that tension is stably transmitted between cells along the entire actomyosin cable. The integration of tensile force along the boundary might explain why increased interfacial tension fails to shrink interfaces and drive cellular rearrangements, as it normally does during germband extension, which precedes A–P boundary formation in segments^13,14. It will be interesting to investigate whether specialized attachment sites integrate the actomyosin cable and how new adhesive connections are established after cytokinesis to maintain the continuity of the cable (Fig. 1b).

An important question that arises from these studies is how the juxtaposition of different cell types triggers actomyosin assembly at interfaces along compartment boundaries. Qualitative differences in adhesion molecules between cell types could participate in this assembly. The feasibility of this model is supported by genetic mosaics for the adhesion molecule Echinoid, in which an actomyosin cable forms between Echinoid non-expressing and expressing cells¹⁵. In vivo imaging of MyoII–GFP after CALI treatment to observe how and where new actomyosin assembly occurs after cells have invaded an adjacent compartment might provide insight into how cell–cell interactions influence actomyosin assembly.

The presence of actomyosin cables at the D–V and A–P wing disc boundaries and the A–P boundary within Drosophila embryonic segments, suggests that compartment boundaries have a more active role in cell sorting than was previously appreciated. It will be interesting to see how other cell sorting events rely on adhesive and contractile activities at boundaries and ultimately to determine how these cellular activities are modulated by transcriptional regulators that specify cell identity.