ESCRT-III Assembly and Cytokinetic Abscission Are Induced by Tension Release in the Intercellular Bridge

See allHide authors and affiliations

Science  29 Mar 2013:
Vol. 339, Issue 6127, pp. 1625-1629
DOI: 10.1126/science.1233866

Making the Final Cut

Abscission, the final separation of two daughter cells, was long thought to be an unimportant step in cytokinesis, triggered merely by the cells pulling strongly enough on the bridge to rupture it. Research over the past 10 years, however, has challenged this notion. Defects in cutting the cytokinetic bridge can lead to the formation of large networks of connected cells or to binucleate cells. Lafaurie-Janvore et al. (p. 1625) now show that the forces postmitotic cells exert on the cytokinetic bridge play an important role in abscission: Surprisingly, increasing the tension in the bridge inhibits abscission, while reducing tension induces abscission. This could provide a sensing mechanism to ensure that daughter cells establish sound connections with their surrounding cells and matrix before detaching from one another.


The last step of cell division, cytokinesis, produces two daughter cells that remain connected by an intercellular bridge. This state often represents the longest stage of the division process. Severing the bridge (abscission) requires a well-described series of molecular events, but the trigger for abscission remains unknown. We found that pulling forces exerted by daughter cells on the intercellular bridge appear to regulate abscission. Counterintuitively, these forces prolonged connection, whereas a release of tension induced abscission. Tension release triggered the assembly of ESCRT-III (endosomal sorting complex required for transport–III), which was followed by membrane fission. This mechanism may allow daughter cells to remain connected until they have settled in their final locations, a process potentially important for tissue organization and morphogenesis.

Long thought to be an unimportant step in cytokinesis (1), abscission is now known to be a complex, tightly regulated process. Recent work has identified several elements required for cutting the cytokinetic bridge (26). These include ESCRT (endosomal sorting complex required for transport), which plays a central role by recruiting the microtubule-severing enzyme spastin (7) and inducing membrane fission (8, 9). Abscission times can vary from 1 hour after anaphase (10) to 3 hours or more (3), even within similar cell lines. We wanted to identify factors that regulate abscission timing and their associated molecular events.

We first assayed the effect of substrate coating and cell density. Abscission time was similar for cells plated on glass and at high density on fibronectin, but it increased with decreasing cell density (Fig. 1, A and C, and movie S1). These results exclude a direct effect of substrate coating but suggest that abscission time is shorter when cells are closely packed.

Fig. 1

Low spatial confinement of daughter cells delays abscission. (A and B) Representative images of daughter cell morphology (visualized by α-tubulin–EGFP signal). Insets show 4× zoom on the intercellular bridge. Scale bar, 10 μm. (C and D) Quantification of abscission time. Each dot represents one daughter cell doublet; means are shown as horizontal bars. Fn., fibronectin. ***P < 0.0001 (Kruskal-Wallis test); n.s., not significant.

To discriminate between effects caused by cell spreading or movement and those caused by cell-cell contact, we plated single cells on disk-shaped fibronectin micropatterns. Abscission time increased with increasing disk size (Fig. 1, B and D, and movie S2). We obtained similar results with thin bar-shaped micropatterns that imposed a given maximal separation on the daughter cells but restrained their spreading area (fig. S1). Thus, abscission time was not modulated by cell-cell contact or cell spreading, but rather by the capacity of daughter cells to move apart from each other after division. Counterintuitively, the more cells could move apart, the longer abscission was delayed. To confirm this, we measured cell separation speed during postmitotic respreading and found that the faster the cells separated, the longer the abscission time (Fig. 2A).

Fig. 2

Tension in the intercellular bridge regulates abscission time. (A) Correlation between initial cell separation and abscission time; n = 126. (B) Separation speed as a function of time (means ± SEM). Green line segments indicate separation speed significantly above 0 (t test); n = 35. (C) Distribution of forces between daughter cells (TFM and laser ablation of the bridge); n = 20. (D) Correlation between bridge tension (retraction speed) and initial separation of daughter cells; n = 31. (E) Bridge tension just prior to abscission (Abs.) compared to randomly chosen bridge (Ctrl.). (F) Abscission time as a function of the contractility of the daughter cell doublet; n = 18. (G) Bridge tension for cells treated by Y27632. (H) Delay between drug treatment and abscission. Significance: [(A), (D), and (F)] P value of Spearman test (linear fit); [(E), (G), and (H)] **P < 0.005, ***P < 0.0001 (t test with Welch correction).

Faster-separating cells displayed delayed or impaired microtubule severing, a phenotype similar to small interfering RNA (siRNA) knockdown phenotype of spastin (11) (figs. S2 and S3 and movies S3 and S4). Thus, it seems that when daughter cells respread after anaphase, they pulled on the cytokinetic bridge as they separated, delaying the severing of microtubules by spastin. This model predicts that cells that either do not pull on the bridge or stop pulling undergo abscission rapidly. Careful examination of bridge morphology and use of specific micropatterns (11) suggested that tensed (straight) bridges were indeed correlated with longer abscission time. Often the two daughter cells transiently moved toward one another, bending the bridge, with abscission following shortly thereafter (fig. S4, C and D). When we quantified this behavior, we found that cells stopped moving apart about 10 min before abscission (Fig. 2B).

To directly measure forces exerted on the connecting bridge, we combined traction force microscopy (TFM) and laser ablation (1113) (fig. S5A). This showed that the bridge was bearing forces in the nanonewton range (1.4 ± 0.2 nN; Fig. 2C) (14). Retraction speed of the bridge after ablation (11, 15) (Fig. 2D, fig. S6, and movie S5) revealed similar forces. The daughter cells’ separation speed and bridge tension were strongly correlated, showing that faster-separating daughter cells were pulling more strongly on their bridges (Fig. 2D). We identified bridges on the brink of abscission by imaging microtubules (fig. S7A). Such bridges were more relaxed than average bridges (Fig. 2E), confirming that bridge tension decreased before abscission. Thus, counterintuitively, tension in the cytokinetic bridge delays abscission.

This contradicts an existing model of abscission, which is based on the measurement of traction forces exerted by cells on their substrate after division (1). We thus performed similar experiments but using TFM, and found that cells displayed a high contractility after division and pulled strongly on the substrate (fig. S5B and movies S6 and S7). Nonetheless, the higher these forces were, the longer the abscission time (Fig. 2F).

Cell contractility might thus be responsible for the high force exerted by daughter cells on their connecting bridge. We decreased the cells’ contractility by treating them with the Rho-associated protein kinase (ROCK) inhibitor Y27632. This strongly reduced both bridge tension (Fig. 2G and fig. S8A) and abscission delay (Fig. 2H). Thus, tension in the cytokinetic bridge could be caused by cell contractility, and reducing bridge tension induced abscission.

One potential source of bridge tension is membrane tension, which can be estimated by pulling a membrane tether with a laser trap (16) (fig. S9, A and B, and movie S8). The tether force was on the order of tens of piconewtons, which corresponds to a contribution of membrane tension to the force applied on the bridge on the order of 0.4 to 0.8 nN (11). Thus, membrane tension could make an important contribution to bridge tension. Membrane tension was low in cases where cells had short abscission times (on glass or treated with Y27632; fig. S9, C and D). It was high when cells had long abscission times (on fibronectin; fig. S9C).

Together, these results suggest the following model: When cells are free to move apart after division, high contractility and motility correlate with a high membrane tension, generating a nanonewton pulling force on the cytokinetic bridge, which delays abscission. When cells stop moving apart, bridge tension drops, triggering abscission. To test our model, we reduced bridge tension by artificially cutting the bridge on one side. We followed microtubules in the remaining half-bridge after laser ablation; the central part was stable, and a cut on the other side was observed shortly thereafter (Fig. 3, B to D, and movie S9), as happens during normal abscission (Fig. 3A).

Fig. 3

Artificial release of bridge tension triggers abscission. (A) Time-lapse sequence of the abscission process. (B and C) Ablation on one side of the bridge (B) and control ablation (C); the leftmost image is before ablation (abl.). (D) Quantification of delays between ablation and the cut on the other side of the bridge (red, n = 18), between the first cut and the second cut in control cells (purple, n = 23), and between the control ablation and the abscission (green, n = 21). ***P < 0.0001 (Kruskal-Wallis test). (E and F) Ablation performed respectively on siRNA spastin–treated cells and siRNA CHMP2A–treated cells. (G) Percentage of cells showing abscission after the bridge ablation with a delay shorter than 60 min (black) or longer than 60 min (dark gray), a slow disassembly of the microtubule bundle (light gray), or no abscission for more than 180 min (white) for siRNA control–treated (n = 17), siRNA spastin–treated (n = 16), and siRNA CHMP2A–treated (n = 20) cells. (H) Evolution of the bundle width after ablation. Scale bars, 10 μm (A), 2 μm [(B), (C), (E), and (F)]; time is minutes after anaphase or after ablation; arrowhead shows microtubule bundle pinching; arrow shows abscission.

To rule out a nonspecific effect of ablation on microtubule disassembly, we depleted key abscission proteins by siRNA: spastin, essential for microtubule severing, and CHMP2A (charged multivesicular body protein 2A), an essential component for ESCRT-III complex assembly (fig. S10A and movie S10). Ablation in spastin-depleted cells led to the characteristic ESCRT-III–dependent pinching of the bridge on the other side, but this was not followed by microtubule severing (Fig. 3, E, G, and H, and fig. S7B). Thus, the induced microtubule severing was spastin-dependent and not nonspecific. Ablation in CHMP2A-depleted cells did not lead to any pinching and was not followed by any microtubule severing (Fig. 3, F to H). This finding was expected, because spastin is targeted to the bridge by ESCRT components (7). Thus, ablation of one side of the bridge induces the regular ESCRT-III–spastin abscission pathway on the other side.

To investigate this point, we tagged CHMP4B, one of the ESCRT-III components that forms the late cone preceding abscission, with green fluorescent protein (GFP). As previously described (8, 9), CHMP4B-GFP was recruited to both sides of the midbody, forming two narrow bands. One of the bands then extended on one side, concomitant with a narrowing of the bridge on that side and with a severing of microtubules (Fig. 4A); a similar process then took place on the other side (movie S11).

Fig. 4

ESCRT-III assembly is regulated in time by forces exerted on the bridge. (A and B) Time-lapse sequence of control bridge cleavage (A) and ablated bridge (B). Leftmost image is before ablation (abl.). (C) Quantification of delays between ablation and formation of the ESCRT-III cone (gray), between ablation and abscission (red), and between control ablation and abscission (green). Late (left) and early (right) stages were quantified. **P < 0.05 (Kruskal-Wallis test). (D) Top: Representative curves of CHMP4B accumulation at the midbody (plain curves, left axis) and microtubule bundle width (dashed curves, right axis) after ablation at early stage (gray) and late stage (black). Bottom: CHMP4B accumulation at the midbody (means ± SEM) before ablation (before), just after ablation (ablation), and 10 or 30 min after ablation (early n = 6, late n = 12). AU, arbitrary units. (E) Time lapse of bridge cleavage in ablated SiRNA-treated cells. (F) The proposed model. Cell membrane (yellow), microtubules (red), and ESCRT-III (green) are shown. (G) Final amount of CHMP4B accumulated at the midbody before abscission. (H) Accumulation rate of CHMP4B at the midbody during cytokinesis as a function of abscission time; n = 18. P value of Spearman test (linear fit) is shown. Scale bars, 2 μm [(A), (B), and (E)]. Time is minutes after anaphase or ablation; arrowhead shows microtubule bundle pinching and formation of the ESCRT-III conical structure; arrow shows abscission.

We also performed siRNA knockdowns of spastin and CHMP2A in these cells and observed the expected phenotypes (fig. S10B) (11). To identify the stage of bridge cleavage triggered by the release of tension, we performed laser ablation experiments at different stages, determined by CHMP4B-GFP localization at the bridge. Ablations performed after CHMP4B-GFP recruitment to the midbody resulted in assembly of the conical structure on the other side 10 min later, concomitant with pinching and microtubule severing (Fig. 4, B and C, and movie S12). Ablations performed on bridges before any CHMP4B-GFP was visible led to transient recruitment of CHMP4B-GFP but no subsequent cone formation or abscission (Fig. 4D).

Ablations in spastin-depleted cells led to the formation of CHMP4B-GFP conical structures and pinching with normal timing, but this was not followed by microtubule severing, whereas ablation in CHMP2A-depleted cells did not trigger any CHMP4B-GFP conical structure assembly nor microtubule severing (Fig. 4E). Thus, tension release after ablation specifically induces the very last step of abscission, the assembly of the ESCRT-III conical structure (Fig. 4F).

Our results have allowed us to identify a major source of variability in abscission time: the forces exerted by daughter cells on the cytokinetic bridge. To respread after mitosis, daughter cells pull on their substrate. Depending on their environment, part of this force is transmitted to the cytokinetic bridge, resulting in delayed ESCRT-III assembly and thus in delayed abscission (fig. S11). We measured separation speed in another cell line showing longer abscission time and found a similar relationship between these two parameters (fig. S12) (11). This suggests that the variability in abscission time reported in different cell lines can be explained simply by differences in postmitotic separation speed.

How do forces acting on the bridge delay ESCRT-III assembly? Membrane tension (17) or local curvature (18) could directly affect ESCRT assembly. Indeed, a recent physical model of ESCRT-III assembly implies that membrane tension could interfere with the process (19). Alternatively, mechanically induced signaling processes might be involved; for example, RhoA and Rac1 regulation are involved both in mechanotransduction (20, 21) and abscission (22).

Other mechanisms involved in the regulation of abscission were reported to rely on the regulation of ESCRT complexes at the midbody (23, 24). Here, we also found that bridge tension delays ESCRT-III assembly. Quantifying CHMP4B-GFP recruitment at the midbody showed that although the final level of accumulation at abscission was not correlated with abscission time, the rate of accumulation was lower in cells in which abscission was delayed (Fig. 4, G and H, and fig. S10C).

We propose that cells possess a sensory mechanism, potentially involving membrane tension, that delays abscission by preventing ESCRT-III accumulation and assembly as long as forces are exerted on the cytokinetic bridge. This mechanism would allow daughter cells to reestablish connections after mitosis—thus preventing tissue rupture or cell loss—and achieve correct relative positioning.

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S14

Movies S1 to S12

References (2547)

References and Notes

  1. See supplementary materials on Science Online.
  2. Acknowledgments: We thank D. Gerlich for the Hela Kyoto α-tubulin-EGFP histone2B-mCherry cell line and the α-tubulin plasmids; I. Poser and A. Hyman for the CHMP4B-GFP cell line; A. Roux, A. Echard, N. Minc, C. Janke, and members of the Piel lab for critical reading of the manuscript and C. Ainsworth for editing it; L. Sengmanivong, V. Fraisier, the Nikon Imaging Center, and the PICT-IBiSA of the Institut Curie for technical support in microscopy; C. Sykes; and E. Terriac and A. Jimenez for helpful discussions. Supported by a French ministerial fellowship (J.L.-J.), a Fondation pour la Recherche Medicale fellowship (J.L.-J. and P.M.), a Ligue Contre le Cancer grant (M.P.), ANR-11-JSV5-0002 (T.B.), and Fondation Nanosciences RTRA (M.B.). The data are contained in the manuscript and the supplementary materials.
View Abstract

Stay Connected to Science

Navigate This Article