Adhesion Functions in Cell Sorting by Mechanically Coupling the Cortices of Adhering Cells

See allHide authors and affiliations

Science  12 Oct 2012:
Vol. 338, Issue 6104, pp. 253-256
DOI: 10.1126/science.1225399

Embryonic Cell Sorting and Movement

Differential cell adhesion has long been thought to drive cell sorting. Maître et al. (p. 253, published online 23 August) show that cell sorting in zebrafish gastrulation is triggered by differences in the ability of cells to modulate cortex tension at cell-cell contacts, thereby controlling contact expansion. Cell adhesion functions in this process by mechanically coupling the cortices of adhering cells at their contacts, allowing cortex tension to control contact expansion. In zebrafish epiboly the enveloping cell layer (EVL)—a surface epithelium formed at the animal pole of the gastrula—gradually spreads over the entire yolk cell to engulf it at the end of gastrulation. Behrndt et al. (p. 257) show that an actomyosin ring connected to the epithelial margin triggers EVL spreading both by contracting around its circumference and by generating a pulling force through resistance against retrograde actomyosin flow.


Differential cell adhesion and cortex tension are thought to drive cell sorting by controlling cell-cell contact formation. Here, we show that cell adhesion and cortex tension have different mechanical functions in controlling progenitor cell-cell contact formation and sorting during zebrafish gastrulation. Cortex tension controls cell-cell contact expansion by modulating interfacial tension at the contact. By contrast, adhesion has little direct function in contact expansion, but instead is needed to mechanically couple the cortices of adhering cells at their contacts, allowing cortex tension to control contact expansion. The coupling function of adhesion is mediated by E-cadherin and limited by the mechanical anchoring of E-cadherin to the cortex. Thus, cell adhesion provides the mechanical scaffold for cell cortex tension to drive cell sorting during gastrulation.

Cell adhesion and cortex tension are commonly assumed to function in cell sorting by controlling cell-cell contact formation (15), with adhesion increasing the contact size and cortex tension decreasing it (37). To clarify how cell adhesion and cortex tension function in progenitor cell-cell contact formation and sorting during zebrafish gastrulation, we first developed a mechanical description of two progenitor cells in contact, on the basis of previous models of cell-cell adhesion and sorting (4, 5). The cells are described as fluid objects with a viscoelastic cortex under tension and adhesive bonds maintaining the cell-cell contact. The size of the cell-cell contact is determined by the balance of forces at the contact boundary:

Embedded Image (1)

where θ is the contact angle of the two adhering cells (Fig. 1A) (8, 9). The tension γi at the cell-cell interface has a positive contribution arising from the cortex tension γcc of the two cells at the contact and a negative contribution arising from adhesion (adhesion tension) of magnitude ω. Outside of the contact, the tension at the cell-medium interface is equal to the cortex tension γcm at this interface.

Fig. 1

Surface tensions and contact shape in progenitor cell doublets. (A) Surface tensions controlling cell doublet shape at steady state. The contact angle θ results from the balance between the adhesion tension ω and the cortex tensions at the cell-medium γcm and cell-cell interfaces γcc (9). (B) Homotypic ectoderm (ecto), mesoderm (meso), and endoderm (endo) doublets during contact formation (movies S1 to S3). Scale bar, 10 μm. Measured contact angles θ are plotted over time as mean ± SEM binned over 100 s for ecto (red, n = 39), meso (green, n = 20), and endo (blue, n = 26) doublets (table S1). (C) Sketch of homotypic triplets before and after separation (9). Bulge volume Vb is measured at the former cell-cell contact after separation and normalized to the cell body volume Vc for ecto (n = 11), meso (n = 13), and endo (n = 5) triplets (movies S4 to S6 and table S2). (D) Tension ratios γcccm and ω/2γcm computed from homotypic triplet and doublet shapes plotted for ecto (n = 11), meso (n = 13), and endo (n = 5). Mean ± SEM (table S3) (9). (E) Sketch of myosin [Myl12.1-eGFP (enhanced green fluorescent protein)] localization in homotypic doublets of pTol2-β-actin::myl12.1-eGFP transgenic zebrafish. Measured mean fluorescence intensity at the cell-cell interface (Icc) is normalized to the mean intensities at the cell-medium interfaces (Icm) of both cells for ecto (n = 25), meso (n = 33), and endo (n = 17) doublets. Scale bar, 5 μm.

To characterize the mechanical parameters that control progenitor cell-cell contact formation, we first determined the ratio of the interfacial tensions at the cell-cell and cell-medium interfaces by measuring the contact angle θ of freely adhering cell doublets ex vivo and using Eq. 1. Homotypic ectoderm doublets showed a larger contact angle θ and consequently a lower ratio of cell-cell to cell-medium interfacial tensions γi/2γcm than mesoderm and endoderm doublets (Fig. 1B).

To derive the relative contribution of the cortex tension γcc and the adhesion tension ω to the ratio of cell-cell to cell-medium interfacial tensions (2γcc ω)/2γcm, we measured the ratio of cortex tensions at the cell-cell to cell-medium interfaces γcccm. To this end, we probed cortex tension at the cell-cell interface by separating contacting progenitor cells ex vivo, using a dual pipette aspiration assay (DPA) and analyzing resulting shape changes directly after separation (Fig. 1C). We observed a rapid increase in curvature in the region of the dissolved cell-cell contact, indicative of reduced cortex tension γcc at this location compared to cortex tension γcm at the cell-medium interface. To calculate the ratio of cortex tensions γcccm from the curvature change at the dissolved cell-cell contact in the different progenitor cell types, we modeled the cell-cell and cell-medium interfaces as elastic shells under tension, consistent with the short time scales of our experiment (9). Using this model, we found γcccm to be lower in contacting ectoderm cells compared to mesoderm and endoderm cells (Fig. 1D). Consistent with this, we observed that in progenitor cell doublets, nonmuscle myosin-2 (Myl12.1) was reduced at cell-cell compared to the cell-medium interfaces, and that this reduction was more pronounced in ectoderm compared to mesoderm and endoderm doublets (Fig. 1E).

Having determined the ratio of tensions at the cell-cell and cell-medium interfaces, we computed the adhesion tension ω using Eq. 1. We found that for all three progenitor cell types, the magnitude of the adhesion tension ω was considerably smaller than the cortex tension γcc at the cell-cell interface (Fig. 1D). This indicates that the cell-cell interfacial tension γi is dominated by the cortex tension γcc at this interface. It further suggests that the cell-cell contact angle θ, and thus the contact size, is predominantly controlled by the ratio of cortex tensions γcccm between these interfaces and that adhesion tension ω, contrary to previous suggestions (8), has only little function in contact expansion.

Although these findings argue against a critical function of adhesion tension in cell-cell contact expansion, formation of adhesive bonds is still essential to mechanically couple the contractile cortices of the adhering cells at the contact and to support stresses normal to the adhesion zone arising from cortical tension, intracellular pressure, or external forces (10, 11). Notably, this mechanical coupling function of adhesion is distinct from the role adhesion plays in providing the adhesion tension ω, where it acts in the direction tangential to the contact zone, thereby expanding it. To characterize the function of mechanical coupling due to adhesion in progenitor cell-cell contact formation and sorting, we asked how its strength is controlled in the different progenitor cell types, and whether interfering with its strength affects progenitor cell-cell contact formation and cell sorting.

To determine the strength of mechanical coupling due to adhesion, we used the DPA assay to mechanically separate homotypic progenitor cell doublets ex vivo and measure the corresponding separation force (12). Considering that cell-cell separation was achieved rapidly (<1 s), our experiments can be described as the separation of two elastic solids, where the separation force Fs depends on different factors, including the number and dissociation rate of adhesion bonds (9, 13). We found that ectoderm doublets exhibit higher Fs at contact times varying from 1 to 10 min than mesoderm and endoderm doublets (Fig. 2A). This difference in Fs between the progenitor cell types is not just a consequence of differences in cell-cell contact size, because normalizing Fs by the contact radius (Fs/Rc) (9) still yielded higher values for ectoderm compared to mesoderm and endoderm doublets (Fig. 2B and fig. S1). Differences in adhesion molecule density at the contact are also unlikely to account for the different Fs, as the expression levels of E-cadherin (Cdh1) and associated proteins, previously shown to mediate adhesion in all three progenitor cell types (1), did not correlate with the different Fs in these cells (fig. S2).

Fig. 2

Contact strength and structure in progenitor cell doublets. (A) Separation force Fs of ectoderm (red; n = 104/41/18), mesoderm (green; n = 30/16/11), and endoderm (blue; n = 44/23/16) homotypic doublets is plotted as mean ± SEM at 1-, 5-, and 10-min contact times (movie S7 and table S4) (9). (B) Ratio of separation force Fs to contact radius Rc of ecto (n = 37), meso (n = 15), and endo (n = 25) homotypic doublets after 5-min contact time (table S5). (C) Sketch of a cell doublet showing cadherin (purple), actin (cyan), and myosin (orange). Optical sections through cell-cell contacts of homotypic doublets stained with antibodies against Cdh1, Ctnnb1, and Ctnna; phalloidin for F-actin; or expressing Myl12.1-mCherry (fig. S3). Scale bar, 5 μm.

We next asked whether differences in the mechanical resistance of adhesion bonds to pulling forces, previously implicated in controlling Fs (12), are responsible for the differences in Fs between the progenitor cell types. To this end, we first determined whether the binding strength of cadherins across the cell-cell contact or to the cortical cytoskeleton limits their mechanical resistance by analyzing the segregation of adhesion complex components during cell-cell separation. Before separation, Cdh1, β-Catenin (Ctnnb1), α-Catenin (Ctnna), and actin, but not Myl12.1, were found to accumulate in a dense ring-like structure at the margin of the cell-cell contact in ectoderm and, to a lesser extent, mesoderm and endoderm doublets (Fig. 2C and figs. S3 and S4). This suggests that higher Fs values coincide with adhesion molecules accumulating at the contact edge. Upon separation by means of the DPA, progenitor cells remained connected via long plasma membrane tethers, and N-cadherin (Cdh2; see below for using Cdh2 as a proxy for Cdh1) together with Ctnnb1, which directly binds to Cdh2 (10), accumulated in these tethers (Fig. 3A). By contrast, Ctnna, which directly or indirectly couples Ctnnb1 to the actin cytoskeleton (14, 15), did not colocalize with Cdh2 and Ctnnb1 in the tethers, but instead disassembled from the dissolving cell-cell contact together with actin (Fig. 3A). These findings indicate that cadherins dissociate from the cytoskeleton during progenitor cell separation, suggesting that cytoskeletal anchoring of cadherins limits the mechanical resistance of adhesion bonds to pulling forces.

Fig. 3

Cytoskeletal anchoring of Cdh2 in progenitor cell-cell contact formation. (A) Ectoderm homotypic doublets expressing Cdh2-eGFP plus LifeAct-RFP (red fluorescent protein) (n = 15), eGFP-Ctnnb1 (n = 20), or Ctnna-eGFP (n = 30) at 5 min contact time before and after separation (movies S8 to S11). Fluorescence intensity at the contact and contact size measured during the separation process (purple and/or cyan on diagram) are plotted as mean ± SEM relative to the value before the separation (table S6). Scale bars, 10 μm. (B and C) Ratio of separation force to contact radius Fs/Rc (B) or contact angle θ (C) measurements of ectoderm (red) or endoderm (blue) homotypic doublets expressing Cdh2-eGFP (n = 17/20) or Cdh2Δcyto-eGFP (n = 20/14) at 5-min contact time. Values are plotted as mean ± SEM and normalized to Cdh2 expression level at the measured contact and at the contact of ecto doublets (movies S12 to S15 and tables S7 and S8).

To explore whether modifying cytoskeletal anchoring of cadherins, and thus the mechanical resistance of adhesion bonds to pulling forces, affects Fs and cell-cell contact formation, we expressed either a full-length version of Cdh2 or a truncated version of Cdh2 that cannot bind to the cortical cytoskeleton (Cdh2Δcyto) (12) in progenitor cells deprived of endogenous Cdh1 expression (cdh1 morphant cells). Expressing similar levels of Cdh2 in homotypic ectoderm and endoderm cell doublets changed neither their endogenous relative difference in Fs normalized by the contact radius Fs/Rc nor their contact angle θ. However, expressing Cdh2Δcyto strongly reduced Fs/Rc and θ to similar low levels in both cell types (Fig. 3, B and C, and figs. S5 and S6). This suggests that proper cytoskeletal anchoring of cadherins is an essential factor determining the difference in separation force and cell-cell contact size between the progenitor cell types.

We next asked how far the observed effect of cadherin cytoskeletal anchoring on Fs/Rc and θ influences progenitor cell sorting. Consistent with a critical function of cell-cell contact formation in cell sorting, replacing endogenous Cdh1 with either full-length or truncated Cdh2 in homotypic ectoderm or endoderm progenitor cell aggregates led to cells expressing Cdh2 being sorted into the middle, surrounded by cells expressing Cdh2Δcyto (Fig. 4A). These observations suggest that proper cytoskeletal anchoring of cadherins is essential for progenitor cell sorting.

Fig. 4

Progenitor cell sorting in vitro and cell-cell contacts structure in vivo. (A) Sorting of red- or green-labeled Cdh2-eGFP– or Cdh2Δcyto-eGFP–expressing progenitors in ectoderm or endoderm cell aggregates (movies S16 to S19 and fig. S7). Scale bar, 10 μm. (B) Myl12.1-eGFP localization within the shield region of pTol2-β-actin::myl12.1-eGFP transgenic zebrafish at 6 hours postfertilization (hpf) (right to the sketch); arrows demarcate the epiblast-hypoblast boundary (epi-hypo, movie S20). Exemplary cells are shown within the lateral mesendoderm (top) and at the germ ring margin (bottom). Red arrows point to Myl12.1 accumulation at the cell-interstitial space interface, and green arrows to My12.1-depleted zones at the cell-cell interfaces (fig. S8). Scale bars, 10 μm. (C) Ring-like accumulation of Ctnnb1 at the contact margin between epiblast cells at the animal pole revealed by antibody staining (movie S21). Boxes highlight the contact on the imaging planes and orthogonal views. Scale bar, 10 μm. (D) Membrane tethers (arrows) formed between separating cells within the lateral mesendoderm in 7-hpf embryos expressing hRas-eGFP (top, movie S22) or within the animal pole of shield-stage embryo expressing Cdh2-eGFP (bottom, movie S23). Scale bars, 10 μm.

To address how our observations of cell doublets ex vivo relate to the situation within the gastrulating embryo, we asked whether progenitor cells segregate in vivo by using the same cell-cell contact formation strategies as their counterparts ex vivo. Analysis of the cell cortex in germ layer progenitor cells in vivo showed that similar to what we observed ex vivo, cortical Myl12.1 was preferentially located at cell-interstitial space/extracellular matrix interfaces (Fig. 4B). This suggests that cortex tension is reduced at cell-cell interfaces both ex vivo and in vivo. Moreover, analysis of progenitor cell-cell contacts in vivo revealed a ring-like localization of adhesion molecules reminiscent of the situation in cell doublets ex vivo (Fig. 4C). Finally, progenitor cells separating in vivo often formed membrane tethers at their dissolving contacts containing Cdh2 (Fig. 4D), similar to cell doublets ex vivo that were separated by means of the DPA. These marked similarities between the situations ex vivo and in vivo suggest that progenitor cell sorting in vivo is driven by the same mechanisms of cell-cell contact formation as observed ex vivo.

The mechanical coupling function of cadherins at the cell-cell contact critically requires sufficient anchoring of cadherins to the cortical cytoskeleton. Considering that mechanical load on cadherin adhesion complexes has previously been suggested to modify its cytoskeletal anchoring strength (15, 16), it is intriguing to speculate that this anchoring strength in the different progenitor cell types is determined by the cortex tension of these cells pulling on the mechanosensitive adhesion complexes, thereby setting their anchoring strength.

Supplementary Materials

Supplementary Text

Materials and Methods

Figs. S1 to S10

Tables S1 to S12


Movies S1 to S23

References and Notes

  1. Information on materials and methods is available on Science Online.
  2. Acknowledgments: We thank M. Biro, J.-Y. Tinevez, and J. Rönsch for technical assistance, V. Barone for critical reading of earlier versions of this manuscript, and the facilities of the MPI-CBG and IST Austria for their continuous support. This work was supported by the Max-Planck-Society and grants from the Polish Ministry of Science and Higher Education to E.P. (454/N-MPG/2009/0), the Human Frontier Science Program to E.P. (RGY0067/2008), and the Fonds zur Förderung der wissenschaftlichen Forschung (FWF) to C.-P.H. (HE3231/6; I812-B12).

Stay Connected to Science

Navigate This Article