Main

First, the differences could be explained by cell-type-specific differences in cytokinesis regulation. We evaluated three human epithelial cell lines2, but the mouse fibroblasts examined by Weaver et al. may generate greater traction forces3 that could promote cytokinesis completion through cell tearing4, rather than by the targeted delivery and fusion of vesicles near the midbody5. Furrow regression occurs many hours after mitosis in HeLa cells2 and, during the intervening period, cells remain connected by a thin cytoplasmic bridge. In fibroblasts, increased traction forces could break this bridge, uncoupling chromosome nondisjunction from furrow regression. Analysis of cytokinesis completion in CENP-E+/+ and CENP-E−/− mouse fibroblasts by time-lapse imaging is necessary to exclude this possibility. Furthermore, it is essential that the CENP-E removal experiment be repeated in epithelial cells, which give rise to most human cancers.

Second, conclusions based solely on the analysis of CENP-E−/− cells is complicated by the fact that CENP-E itself (and its binding partner, BubR1) may be involved in coupling chromosome nondisjunction and cytokinesis completion. CENP-E is required for efficient congression of mono-orientated chromosomes to the metaphase plate6 and for spindle checkpoint signalling7. Following anaphase, CENP-E localizes to the midbody8 where it may negatively regulate cytokinesis completion9. Thus, given the many important functions of CENP-E in mitosis, CENP-E could act directly in coupling nondisjunction to cytokinesis completion, explaining why CENP-E removal would uncouple these two processes.

Third, nondisjunction in CENP-E−/− cells is unlike spontaneous nondisjunction. Nondisjunction induced by loss of CENP-E results from cells that enter anaphase with chromosomes remaining at the poles10. In contrast, we identified 17 cells that underwent nondisjunction, yet only two cells showed defects in chromosome congression2. Spontaneous nondisjunction therefore seems to be a different process from nondisjunction occurring in CENP-E−/− cells, providing another potential explanation for differences in the subsequent regulation of cytokinesis. The causes of spontaneous nondisjunction are poorly understood, but our time-lapse analysis indicates that perturbation of CENP-E function is unlikely to be among them.

Weaver et al. suggest that furrow regression is due to the presence of DNA in the cleavage furrow1, but this possibility is inconsistent with our data: of 395 cells that completed cytokinesis, 58 showed chromosome bridging or lagging, indicating that these defects are not sufficient to induce furrow regression2. Furthermore, 47% of cells that became binucleated showed no bridging or lagging, indicating that these defects are not required for furrow regression. Although thin strands of trapped chromatin may be beyond our limit of detection, these cells also showed no evidence of micronuclei or disrupted interphase nuclear architecture, which might arise as a result of such defects. In contrast, cytokinesis failure has been reported only in cells showing gross distortions in nuclear architecture, with teardrop-shaped nuclei connected to one another by a large chromatin bridge11. We observed defects of this magnitude in only 11% of cells that became binucleated. Thus, although bridging and lagging may be associated with binucleation in a fraction of cells, we conclude that other mechanisms must exist to link nondisjunction and furrow regression in the remaining cells.

In budding yeast, there is a pathway that delays cytokinesis completion to prevent damage to chromosomes near the division site12. Although this mechanism has not been demonstrated in animal cells, our results indicate that such a pathway could possibly be activated more generally by spontaneous nondisjunction.