INTRODUCTION Cytokinesis of animal cells begins with a constriction mediated by an actomyosin contractile apparatus. When this constriction has narrowed maximally, the daughter cells can either move apart and separate, as fibroblasts do, or they can cohere, as early embryonic cells do. The latter way of completing cytokinesis has been studied for many years in mammalian embryos, especially in the mouse, and is termed compaction. It is well established that compaction is dependent on E- cadherin (Shirayoshi et al., 1983): the cadherin complex is active, i.e. cytoskeleton linked, at regions of cell-cell contact (Ohsugi et al., 1996), and in the presence of antibodies to this cell adhesion molecule, blastomeres remain rounded instead of cohering (Hyafil et al., 1980). To understand how cells cohere after division, it is thus important to learn how the cadherin complex is localized and activated in the right region of daughter cells. In the mouse embryo, where cadherin is initially present on the entire blastomere surface (Vestweber et al., 1987), this may be achieved by local dephosphorylation of tyrosine residues in pre-existing complexes (Ohsugi et al., 1996). In the 1970s and 80s, there was some interest in the role of the cytoskeleton, particularly of microtubules, in the cohesion of mouse blastomeres after division. One reason for this interest is that microtubule arrays had been reported to be present in large numbers adjacent to the surface mediating cohesion (Ducibella and Anderson, 1975). Further EM and conventional fluorescence microscopy studies, however, failed to detect such microtubules (Houliston et al., 1987). Instead microtubules were seen near the apical surface, and were thus proposed to be involved in the polarization of cells during compaction, rather than in mediating cohesion. Evidence that microtubules might have a role in compaction has come from experiments using drugs which perturb microtubules, but such studies have been controversial. Several early studies (Ducibella and Anderson, 1975; Ducibella, 1982), finding that microtubule depolymerization at the 8-cell stage (when compaction occurs) did not prevent flattening, concluded that microtubules are not required. Another study, however, found that nocodazole treatment at the 8-cell stage accelerates flattening, while taxol reverses it (Maro and Pickering, 1984). This led to the proposal that microtubules have a role in constraining cell shape during compaction. Interestingly, treatment of embryos at earlier stages, namely the 2- (Surani et al., 1980) or 4-cell (Ducibella, 1982) stages, led to a failure of compaction. Additionally, these cells could not serve as substrata for the spreading of 8-cell stage blastomeres (Ducibella, 1982). These experiments seem to implicate microtubules, at least prior to the 8-cell stage, in the production of a surface suitable for cohesion. However, these results have to be interpreted with caution because the treatments used led to mitotic arrest, which causes cells to lose their adhesiveness and round up. Hence, it is still not clear whether or not 3695 Journal of Cell Science 111, 3695-3703 (1998) Printed in Great Britain © The Company of Biologists Limited 1998 JCS3864 During the first few cleavages of the zebrafish embryo, daughter blastomeres are loosely associated immediately after furrow ingression, but then gradually cohere. Cohesion appears to be cadherin-dependent, as cadherin and β-catenin are found at the membrane between cohering blastomeres, and blastomeres fail to cohere in calcium-free medium. Cadherin and β-catenin are not initially found on the blastomere surface, but are deposited specifically at the furrow surface. An array of parallel microtubules is present on either side of the furrow tip during ingression, as seen by confocal microscopy of α- tubulin labelled embryos. Transient incubation of embryos in 1 μg/ml nocodazole at the start of furrowing, which causes a loss of the furrow array, inhibits the localization of β-catenin to the furrow surface but does not prevent furrow ingression. During ingression, intracellular membranes are transported to the furrow, as shown by labelling with DiD or DiOC 6 (3). Concentration of these membranes near the furrow surface is microtubule- dependent. These findings suggest that microtubules are required for cohesion of blastomeres because they mediate trafficking of intracellular membranes to the furrow surface, where they are exocytosed and allow cohesion via cadherins. Key words: Cadherin, β-Catenin, Exocytosis, Microtubule, Cytokinesis, Compaction SUMMARY Furrow-associated microtubule arrays are required for the cohesion of zebrafish blastomeres following cytokinesis Suresh Jesuthasan Max-Planck-Institut für Entwicklungsbiologie, Spemannstr. 35/I, 72076 Tübingen, Germany (e-mail: suresh.jesuthasan@tuebingen.mpg.de) Accepted 12 October; published on WWW 18 November 1998