Cytoplasmic flow is a cell size sensor that scales anaphase

Main

Chromosome separation during anaphase is tightly coordinated with nuclear envelope reformation (NER) so that NER occurs after complete segregation of the DNA yet before separating chromosomes reach the edge of the cell (cell pole). NER positioning is regulated by a mechanism in which Aurora B kinase, localized at the spindle midzone, and PP1/PP2A phosphatases act on chromosomal substrates^1,2,3. After substrates are phosphorylated at the midzone, their phosphorylation levels decay exponentially as a function of time. NER occurs only at a particular time, below a defined phosphorylation threshold and, because chromosomes are moving while being dephosphorylated, at a defined distance from the midzone¹. Therefore, (1) the time of NER is set by the dephosphorylation rate and (2) the position of NER is set by the dephosphorylation rate and the velocity with which chromosomes separate from the midzone. NER is positioned according to a temporal model (dephosphorylation rate), which becomes a spatial model by chromosome velocity.

Results

Position of NER scales with cell size

Can this mechanism adjust the position of NER to cell size? We studied this during early zebrafish embryogenesis, where cells undergo nine rapid mitotic cycles (~10 min each) without interphase⁴. In these cleavage divisions, cell length along the axis of chromosome separation decreases by around tenfold (Extended Data Fig. 1a–c). In the one-cell stage, cells are 580 ± 52 µm long and the distance between the two sets of separating chromosomes when NER occurs is 91 ± 14 µm, which is longer than the cell length at the 512-cell stage (68 ± 11 µm). Therefore, the NER position must adapt to cell size.

To study how NER adapts to cell size, we systematically measured NER timing and position during anaphase from 4- to 512-cell stage, ranging from 300 μm to 70 μm in cell length. First, we used fluorescent wheat germ agglutinin (WGA-640) to label glycosylated nucleoporins, the first factors known to be recruited to the chromosomes during anaphase^5,6,7 (Fig. 1a,b). To then monitor NER, when the nuclear envelope is sealed and has a functional nuclear transport system, we followed the nuclear accumulation of green fluorescent protein (GFP)-NLS (where NLS is the nuclear localization signal) (Fig. 1a,b). The timings of WGA recruitment and NER do not change with cell size: WGA is recruited 113 ± 3 s after anaphase onset, and NER occurs at 271 ± 18 s (Fig. 1c). The timing of NER was previously shown to be dependent on Cdk1 downregulation in Drosophila and human culture cells⁸. However, in zebrafish, Cdk1 inhibition did not affect the timings of WGA or NER (Extended Data Fig. 1d–f). We noticed that WGA timing coincides with the transition from anaphase A to B (Fig. 1a,b with Extended Data Fig. 2b). Indeed, like in other systems⁹, here anaphase can be subdivided: in anaphase A, chromosomes approach the spindle poles, while during anaphase B, both chromosomes and spindle poles move together, with the same velocity (Extended Data Fig. 2a–c).

**Fig. 1: Position of NER scales with cell size through scaling of anaphase chromosome velocity.**

Whereas NER timing is independent of cell size, NER positioning is not. WGA recruitment occurs at approximately the same distance regardless of size, but NER occurs at distances proportional to cell size, that is, NER scales (Fig. 1d–f; note that WGA scaling is subtle). We wondered whether either the chromosome phosphorylation threshold for NER or the dephosphorylation rate can explain NER scaling. To address this, we imaged live embryos injected with a fluorescently conjugated antibody that recognizes phosphorylated Histone H3 (pH3-s10)¹⁰, a bona fide substrate of Aurora B. pH3-s10 levels correspond to the phosphorylation state of chromosomes as they move away from the midzone^3,11 (Extended Data Fig. 2d–f).

We first confirmed in this system that the phosphorylation levels decay exponentially with time¹ (Fig. 1g). The dephosphorylation rate, which determines the decay time, is the same regardless of size (Fig. 1g,h). In addition, both WGA recruitment and NER occur at defined thresholds of phosphorylation levels regardless of cell size (75 ± 5% and 37 ± 6% of the midzone level, respectively; Fig. 1i). Because neither the dephosphorylation rate nor the phosphorylation threshold for NER changes, chromosome velocity must be cell size dependent to explain scaling.

During anaphase A and B, chromosomes move at two different velocities, v_A and v_B, respectively. In anaphase A, v_A remains largely constant and shows marginal scaling in the smallest cells (128-cell stage onwards; Fig. 1j–m). By contrast, in anaphase B, v_B strongly scales for the full range of cell sizes (Fig. 1j–m). This explains why the position of WGA recruitment is constant, whereas NER position scales. The scaling of v_B underlies the scaling of NER positioning, prompting us to study what mediates chromosome separation during anaphase B.

Cytoplasmic flows emerge in anaphase

During anaphase A, chromosomes approach the spindle poles and the spindle poles do not move (Extended Data Fig. 2a–c). We showed above that, during anaphase B, the spindle and the chromosomes move together and the chromosomes do not anymore approach the spindle poles (Extended Data Fig. 2a–c), indicating that the spindle is not used to separate chromosomes in anaphase B. Furthermore, astral microtubules are not in direct contact with the cell cortex during anaphase B, excluding pulling forces from the cortex (Extended Data Fig. 2g–i). What then drives chromosome separation in anaphase B? Cytoplasmic flows were shown to transport cytoplasmic contents over large distances^12,13 or to position cellular structures such as the mitotic spindle in mouse oocytes or nuclei in Drosophila embryos^14,15.

We noticed, by looking at mitochondria in early zebrafish embryos, that flows appear during anaphase. While it is well established that mitochondria are moved by motors on microtubules, a collective movement resembling cytoplasmic flows was indeed prominent (Fig. 2a). Thus, we imaged mitochondria in transgenic embryos expressing a mitochondrial targeting peptide tagged with GFP¹⁶. In anaphase, mitochondria are dispersed in the cytoplasm (Extended Data Fig. 3a–d) and can be used to analyse cytoplasmic flows. Particle imaging velocimetry (PIV) analysis revealed flow patterns with flow velocity maximal along the axis of chromosome separation (velocity; Fig. 2b) and vortexes around the anaphase spindle (vorticity; Fig. 2c). Flows emerge at anaphase onset and become stronger in anaphase B (Fig. 2d and Supplementary Video 1).

**Fig. 2: Cytoplasmic flows emerge during anaphase B and are independent of chromosome movement.**

To study whether other organelles are also moved by cytoplasmic flows, we monitored endogenous lipid droplets (Extended Data Fig. 3e–k). We also studied injected lipid droplets, which lack motor proteins and must drift following the flows (Extended Data Fig. 3l,m). In both cases, motility corresponds to the mitochondria flow patterns, validating mitochondria movement as a proxy for cytoplasmic flows. This also indicates that organelles, such as mitochondria and endogenous lipid droplets, or other objects, such as exogenous lipid droplets, are displaced in the cytoplasm together with the local flows, raising the possibility that chromosomes could, too.

In anaphase B, the speed and orientation of the mitochondria flow field adjacent to chromosomes (Methods) matched those of chromosomes themselves (v_flow = 0.08 ± 0.02 µm s⁻¹ versus v_chromo = 0.11 ± 0.04 µm s⁻¹; Fig. 2e and Extended Data Fig. 3n–s). Also, chromosomes move parallel to neighbouring lipid droplets (Extended Data Fig. 3t–v). These neighbouring lipid droplets lack dedicated structures, like the centromere on chromosomes, that could bind to a specific set of microtubules. Since droplets near the chromosomes and chromosomes themselves move with parallel trajectories, flows could move chromosomes. However, flows could themselves be produced by the active movement of chromosomes in a passive fluid. This creates a conundrum: does chromosome movement generate flows, or do flows move chromosomes?

Microtubules are essential for cytoplasmic flows

We addressed this by studying cells dividing without DNA during cleavage divisions, which we observed occasionally (n = 7). The origin of this phenomenon is currently not understood but has been reported before^17,18,19. To follow mitosis in the absence of DNA in these cells, one-cell-stage embryos were injected with fluorescently tagged Histone H1 and α-tubulin to monitor chromosomes and microtubules. Cells without DNA did not form a mitotic spindle but generated two asters during anaphase (Fig. 2f and Supplementary Video 2). The absence of DNA was not due to failure of labelling since neighbouring cells, used as internal controls, have chromosomes and mitotic spindles (Fig. 2f).

Cells without chromosomes progress through mitosis, undergo cytokinesis and produce daughter cells through at least four cell cycles (two cell cycles shown in Extended Data Fig. 4). Remarkably, they show cytoplasmic flows and vortexes like cells with chromosomes (Fig. 2b,c,g–i and Supplementary Video 3). Therefore, cytoplasmic flows are not a passive consequence of the displacement of chromosomes.

Cytoplasmic flows have been shown to be mediated by active mechanisms such as actin polymerization, contraction of the cortical actomyosin network or the movement of molecular motors on microtubules^12,13,14,15. We first looked at actin. From anaphase onset onwards, bulk actin progressively depolymerizes in the cytoplasm and remains mostly at the cell cortex (Extended Data Fig. 5a), as previously reported^20,21. In this scenario, a decreased concentration of bulk actin might not contribute substantially to the generation of cytoplasmic flows in the centre of the cell. Complete actin depolymerization causes loss of embryo shape and ultimately leads to cytoplasmic leakage outside the cell. Thus, we performed a short incubation with latrunculin B, which only partially affected cortical actin while keeping embryo shape but completely inhibited cytoplasmic actin polymerization (Extended Data Fig. 5b). Under these conditions, the flow pattern is maintained (Fig. 3a–c,e–h), but the timing of the flows is changed: flows are initiated and slowed down earlier compared with control embryos (Fig. 3d). This is explained by the fact that, in control embryos, actin depolymerization might fluidize the cytoplasm, allowing flows to emerge mostly during anaphase B (Extended Data Fig. 5a). Upon the downregulation of actin, cytoplasmic actin is already reduced in anaphase A, allowing flows to emerge earlier. Importantly, changes in flow dynamics upon actin downregulation did not affect the position and scaling of NER (Extended Data Fig. 5c–f). These data exclude cytoplasmic actin filaments as the origin of the flows and suggest that flows are rather downregulated by the existence of an actin network.

**Fig. 3: Anaphase astral microtubules are required for the anaphase flows.**

We next focused on the role of astral microtubules in anaphase B. To visualize the displacement within the microtubule network, we performed tubulin speckle imaging, in which the microtubule lattice is stochastically labelled^4,22,23. Speckles showed marked flows during anaphase, with velocity and vortexes like those observed with labelled mitochondria (Fig. 3i–m,p). These results are consistent with the idea that the displacement of astral microtubules generates the anaphase cytoplasmic flows.

To study whether microtubules themselves generate flows, we used a photo-activatable microtubule depolymerizing drug (SBTub3-P)²⁴ at the metaphase–anaphase transition. Under these conditions, the mitotic spindle architecture is maintained and chromosome separation in anaphase A was only mildly affected (WGA position is 37.9 ± 3.5 and 30.9 ± 5.4 µm in control and SBTub3-P, respectively; four-cell stage) (Fig. 3q and Extended Data Fig. 6a–d). However, during anaphase B, astral microtubule growth was impaired (Extended Data Fig. 6a–d), the mitochondrial flow showed no vortexes and chromosome velocity was reduced (Fig. 3r–u and Extended Data Fig. 6e). Under these conditions, the actomyosin network should not be affected. This suggests that microtubules are essential for cytoplasmic flows, while the cortical actin remaining in the latrunculin experiment above (Fig. 3a–h and Extended Data Fig. 5b) does not contribute to the generation of flows. Instead, the latrunculin experiment suggests that the actin network is a facilitator for flows to emerge.

Drag of motor-driven organelles generates cytoplasmic flows

While tubulin speckles show flows with vortexes, microtubule growth, monitored by EB3-labelled plus-ends, is radial, away from the aster centre (Fig. 3n–p). Microtubule growth occurs on a different (faster) timescale than the flow of speckles or mitochondria. The difference between the pattern in the flows (vortexes) and growth (radial) is explained because flows are too slow to have a major impact on the short-lived growth events monitored by EB3. Indeed, microtubule polymerization cannot displace the cytoplasm and induce flows by itself, because polymerization merely represents the reallocation of the tubulin heterodimer mass into the microtubule filament. Instead, the movement on microtubules of bulky organelles fuelled by motors could generate flows.

Dynein transports cargo, including bulky organelles (mitochondria, endoplasmic reticulum and others), towards the minus ends of microtubules that are concentrated in the aster centre^25,26. The movement of bulky cargo against a viscous cytoplasm causes drag on the cargo, which in turn could displace microtubules in the opposite direction²⁷. Indeed, dynein was shown to generate large cytoplasmic pulling forces in sea urchin embryos^28,29 and Xenopus laevis extracts³⁰. We therefore quantitatively analysed the minus-end motion of dynein with mitochondria, as an example of a bulky cargo. By high-temporal-resolution imaging, we tracked single fluorescently labelled mitochondria while moving into a cleared bleached area near the aster centre (Fig. 4a). The analysis of mitochondrial motility, including mean square displacement, shows two types of episodes within tracks in anaphase (Fig. 4b–f and Extended Data Fig. 7): (1) fast directed motion towards the aster centre (v = 0.95 ± 0.3 µm s⁻¹; fast episodes) and (2) a combination of diffusive (D = 0.017 ± 0.02 µm² s⁻¹) and slow directed motion (v = 0.089 ± 0.05 µm s⁻¹; slow episodes) with the same direction and velocity as cytoplasmic flows. Therefore, mitochondria can be either engaged on microtubules by dynein and moving towards the aster centre (fast episodes) or disengaged from dynein (slow episodes) and drifting with the cytoplasmic flows.

**Fig. 4: Friction of bulky cargo bound to dynein against a viscous cytoplasm generates the anaphase flows.**

To test the role of dynein itself, we injected p150-CC1, a dominant-negative fragment of dynactin that inhibits dynein function^31,32. In these conditions, the spindle architecture was maintained, as previously reported³³. During anaphase A, chromosome velocity was not affected (Fig. 4k). However, in anaphase B, fast mitochondrial episodes were reduced, flow patterns were abolished and chromosome velocity was strongly affected (Fig. 4g–n). Therefore, dynein generates flows in the microtubule network by moving bulky cargo.

We then studied theoretically whether the transport of a cargo (mitochondria) by motors (dynein) in the presence of friction drag can generate a considerable movement of a microtubule in the opposite direction (Supplementary Information section 1.1-3). Theory shows that the ratio of the velocities of mitochondria and microtubule is equal to the inverse of the ratio of their friction coefficients (Supplementary Information section 1.1). Friction depends on the size and shape of the object. Considering the (longitudinal) friction coefficient³⁴ of a 10-µm-long microtubule⁴ and that of a mitochondrion with 1 µm diameter, which are both similar (({xi }_{text{mt},parallel }approx {xi }_{text{mito}}approx 400,{{mathrm{pN}}}/(upmu {mathrm{m}},{mathrm{s}}^{-1}))), the two velocities are comparable. Therefore, the drag of one mitochondrion is considerable and can contribute to the collective displacement of the microtubule network. Other organelles could also contribute to the observed flows.

Cargo drag as described above occurs in a scenario with a single aster. In this configuration, microtubules would be displaced away from the aster centre with radial symmetry and the aster centre would remain in place. Instead, experimentally, we observed a flow pattern where the asters, with the chromosomes therein, move towards the cell poles. This is explained by the fact that astral microtubules do not extend beyond the midzone, creating an asymmetric aster (Extended Data Fig. 6f,g and Supplementary Fig. 1). In turn, astral asymmetry causes an imbalance in cargo drag that ultimately leads to polar movement of each aster (Supplementary Information section 1.3). In summary, dynein-dependent motility with friction drag on a cargo, in the context of aster asymmetry, can generate flows that separate chromosomes in anaphase. What is then causing scaling of these flows?

Confinement scales cytoplasmic flows

NER scaling is mediated by the scaling of velocity of chromosome motion, which is indeed mediated by cytoplasmic flows. We therefore studied whether flows scale as cells become smaller. Flow velocity adjacent to chromosomes does scale (Fig. 5a) and correlates with chromosome velocity in cells of different sizes (Fig. 5b and Extended Data Fig. 8a–g). Importantly, flow velocity in other regions (for example, spindle midzone) correlated neither with chromosome velocity nor orientation (Extended Data Fig. 8h,i).

**Fig. 5: Anaphase cytoplasmic flows scale by confinement that ultimately scales NER.**

Which parameter scales to mediate flow scaling? Centrosome size and microtubule dynamics were shown to scale in metaphase spindles^35,36. We measured microtubule density, microtubule growth, aster size and aster growth velocity, in anaphase; none of these experienced a major change in cells of different sizes (Extended Data Fig. 9a–d) and neither did the density of mitochondria nor the proportion of mitochondria, which gets engaged on dynein-dependent cargo change (fluorescence recovery after photobleaching (FRAP) analysis; Extended Data Fig. 9e,f). Finally, velocity, run length and processivity of mitochondria engaged with microtubules are the same regardless of cell size (Extended Data Fig. 9g–i). Thus, changes in these biochemical features underlying cytoplasmic flows are not changed during scaling.

Could flows scale by the boundary effects on the aster, whose size does not scale, in cells of decreasing size (confinement)? We tested this with a numerical study (Supplementary Information section 2) in which we considered parameters based on values obtained experimentally from our system and other cellular systems (Supplementary Table 1).

In confinement conditions (Fig. 5c–h), the polar displacement of asters led to a flow pattern that moves chromosomes (Fig. 5c), forms vortexes (Fig. 5e,g) and scales with cell size (Fig. 5f,h), so that NER position itself scales (Fig. 5d), as observed experimentally (Fig. 5a,d). Considering the full range of cell sizes of the zebrafish cleavage-stage embryos (from ~600 µm in the 1-cell stage to 60 µm in the 512-cell stage), the theoretical analysis yields, as an emerging property, the appearance of an upper limit, where scaling reaches a plateau (Fig. 5d). Such a plateau was described before for metaphase spindle scaling in zebrafish and Xenopus^4,37. Indeed, for larger cells (one- and two-cell stage), the observed NER position behaves according to that upper limit. In other words, our theoretical framework based on confinement captures in quantitative detail the relationship between cell size and NER position, not only when it scales, but also when it does not anymore (Fig. 5d).

How does confinement slow down flows? The movement of an object in a confined system decreases as it gets closer to the boundary (no-slip boundary effect)³⁸. Therefore, flows experience a larger drag and become slower the closer they are to the boundary. This effect is more prominent in small than in large cells, explaining flow scaling. For the largest cells, however, the distance to the cell boundary is too large to considerably affect the motion of the asters, thereby explaining the upper limit.

Taken together our experimental and theoretical approaches uncover a scenario where confinement by itself causes scaling of flows and, ultimately, determines NER position. To further test whether the mere size of cells determines the position of NER, we generated smaller-sized embryos by yolk aspiration in one-cell stage, before streaming of ooplasm from the yolk towards the cell²⁰. We thereby achieved a systematic cell size reduction in two- and four-cell stages, compared with control embryos (Extended Data Fig. 9j), without significantly affecting any of the parameters that contribute to NER (Fig. 5i and Extended Data Fig. 9k–q). In a scenario where a limiting biochemical factor decays with time during cleavage stages and is essential for scaling, we expect NER positioning to be determined by timing (developmental stages) rather than by cell size. Conversely, if scaling is defined by confinement, NER position would be determined by the new cell size in the aspiration experiment, not by the stage. We observed that NER is positioned according to cell size rather than stage (Fig. 5j–l), indicating that confinement drives NER scaling; the aspiration data fall into the control scaling curve of NER distance as a function of cell size (Fig. 5k).

To test which cell size parameter (volume versus cell length) has a stronger impact on NER positioning, we patterned two-cell-stage embryos in well-defined rectangular agarose templates, achieving an inversion of the cell aspect ratio (Extended Data Fig. 10a,b). Under these conditions, the flow pattern was maintained: flows were aligned with the axis of chromosome separation, and vortexes were formed in the vicinity of the spindle region (compare Fig. 2a–c with Extended Data Fig. 10d–f). Upon insertion into the agarose template, cells in two-cell-stage patterned embryos have the same volume as cells in control embryos, while their length halves, matching that of cells in eight-cell-stage control embryos. Extended Data Fig. 10c shows that NER positioning occurs according to cell length rather than cell volume.

Discussion

Here, we identified cytoplasmic flows as a mechanism for chromosome separation in anaphase that provides cell size information and mediates scaling of the position of NER. This mechanism is supported by the following five key findings: (1) the position of NER scales with cell size (Fig. 1d–f); (2) NER scales because chromosome velocity scales (Fig. 1j,k); (3) chromosomes are moved by cytoplasmic flows that emerge in anaphase (Fig. 2a–d); (4) flows are generated by the friction drag of bulky cargo moved by dynein when walking on microtubules, against a viscous cytoplasm (Fig. 4b,c); and (5) flows scale by their physical confinement in cells of decreasing sizes (Fig. 5a,d,k). Cytoplasmic flows in confinement are sufficient to scale chromosome velocity and, consequently, the position of NER.

During anaphase, the position and timing of NER are coupled by chromosome velocity. In our study, we found that the timing of NER, determined by the biochemistry of the dephosphorylation rate of chromosomal substrates, does not itself scale (Fig. 1g–i). By contrast, NER position does scale because the scaling of velocity, mediated by flows, brings chromosomes further away in the same time interval. The scaling of velocity cannot be explained by changes in biochemical properties (Extended Data Fig. 9a–i). Instead, flows scale simply by the physics of confinement. Because flows, in turn, affect chromosome velocity and the position of NER, flows become a ‘sensor’ of the confinement state and, therefore, of the cell size used for scaling. Our data raise the possibility that cytoplasmic flows could be a general, simple mechanism for the scaling of other cellular structures or processes within the cytoplasm, such as spindle positioning^14,39,40, nuclear spacing or the length of the morphogen gradient in the syncytial blastoderm of flies^15,41.

Methods

Zebrafish lines and maintenance

This study followed European Union directives (2010/63/EU), the Swiss Animal Protection Act and the Swiss Animal Welfare Ordinance. Zebrafish lines used were maintained in a recirculating system with a 14 h day and 10 h night cycle at 28 °C. To visualize chromosomes, the Tg(h2afva:h2afva-GFP)⁴² and Tg(Ef1α:H2B-mCherry) transgenic lines were used. Mitochondria were visualized with a Tg(Ef1α:MLS-GFP) transgenic line¹⁶. F-actin lines, Tg(actb1:Utr-GFP) and Tg(actb1:Utr-mCherry), were a gift from the laboratory of Carl-Philip Heisenberg²⁰. Microtubules were visualized either with a transgenic line expressing the microtubule binding domain of Ensconsin tagged with 3 GFPs Tg(bactin2:HsENSCONSIN17-282-3xEGFP), a gift from Martin Wuhr³², or with a transgenic line expressing human Doublecortin, Tg(actb2:EGFP-Has.DCX)⁴³. The AB wild-type strain was used for the injection of labelled markers.

Live imaging

For all live imaging experiments, embryos from 30 mpf (minutes post fertilization) to 3 hpf (hours post fertilization) were manually dechorionated, mounted on 0.7–1.0% low-melting agarose and maintained at 28 °C in a temperature-controlled chamber. Control embryos that, while imaging, showed errors during mitosis were excluded from subsequent analysis.

Injected markers and drugs

All makers were injected in the yolk of one-cell-stage embryos. Early NER was visualized with WGA conjugated with a 640 CF®dye (29026; Biotium). Complete NER was observed with NLS-GFP (a gift from the Jan Brugués laboratory, MPI-CBG, Dresden). Histone H1 protein purified from calf thymus and conjugated with Alexa Fluor 488 (H1-488) was used to visualize chromosomes (H13188; Thermo Fisher Scientific). The Aurora B phosphorylation gradient was observed using a Fab antibody against phosphorylation at S10 of Histone H3 conjugated with Alexa-488, Alexa-Cy3 or Alexa-Cy5 (injected a 1:10 dilution from 0.2 μg ml⁻¹ stock). Fab antibodies were a kind gift from the Hiroshi Kimura laboratory, Tokyo Tech, Japan¹⁰. Microtubules were visualized with purified/conjugated α-tubulin-ATTO-565, and microtubule speckles were visualized with a 20:80 ratio of labelled:non-labelled mixture of α-tubulin-ATTO-565. Microtubule plus tips were visualized with purified EB3–GFP or EB3–mCherry. Both tubulin and EB3 purified proteins were a kind gift from the Charlotte Aumeier laboratory (University of Geneva, Switzerland). Endogenous lipid droplets were labelled with Nile Red following previous protocols²⁰. SBTub3-P was provided by the laboratory of Oliver Thorn-Seshold (LMU, Munich). P150 purified protein was a kind gift from the Jan Brugués laboratory (MPI-CBG, Dresden).

NER scaling description

The scaling of the time and position of NER in control, latrunculin B, dinaciclib and aspirated embryos was based on live imaging data of Tg(Ef1α: H2B-mCherry) embryos, co-injected with NLS-GFP and WGA-640. Live imaging was performed on a 3i Marianas confocal spinning disk set-up based on a Zeiss Z1 stand and a Yokogawa X1 spinning disk head. Images were acquired with sequential excitation with 488, 561 and 640 nm laser lights, every 20 s, with a Zeiss LD C-APO 40×/1.1 W Korr M27 objective and a z-step of 1.5 μm to a total z-stack of ~25 μm, in the centre of the cell, where the mitotic spindle is positioned. For each mitotic cycle, the centre of the z-stack was adjusted to the position of the spindle.

Live imaging of pH3-s10

Imaging of pH3-s10 was performed in the background of Tg(Ef1α: H2B-mCherry) transgenic embryos. Fab antibody against pH3-s10 conjugated with Alexa-488 was injected in one-cell-stage embryos combined with WGA-640. For the correlation between timing of NER and pH3 levels, pH3-s10-Alexa-Cy5 and NLS-GFP were used. Imaging was performed with the same set-up as for the scaling description.

Live imaging of mitochondria for PIV analysis

Live imaging was performed on Tg(Ef1α:MLS-GFP) transgenic embryos co-injected with WGA-640. When mentioned, H1-488 was co-injected, to visualize chromosomes. WGA-640 signal was used to time the beginning of anaphase B and the time window to analyse the flows. The imaging set-up was the same as for the scaling description, with a time resolution of 10 s and a z-stack of 17 μm. The same conditions apply for control, SBTub3-P, p150 and latrunculin B experiments.

Individual mitochondria tracks

The imaging of individual mitochondria tracks was based on Tg(Ef1α:MLS-GFP) transgenic embryos injected with H1-488 to facilitate the staging of mitosis (metaphase versus anaphase). When mentioned, α-tubulin-ATTO-565 was co-injected to visualize the mitotic spindle. Single z-stack movies were acquired in a Zeiss LSM 780 with a Zeiss C-Apochromat 40×/1.2 W Korr FCS M27, 5× zoom with 0.6 s time interval. FRAP was performed with a square region of interest (ROI) positioned such that a portion or half of the aster region was bleached.

Tubulin speckle microscopy

For tubulin speckle microscopy, wild-type embryos were co-injected at the one-cell stage with α-tubulin-ATTO-565 in a 20:80 ratio of labelled:non-labelled and EB3–GFP. Single z-stack movies were acquired in a Zeiss LSM 780 in Airyscan mode, with a Zeiss C-Apochromat 40×/1.2 W Korr FCS M27, 4× zoom and 4 s time interval.

Microtubule growth with EB3

Wild-type embryos were injected at the one-cell stage with EB3–mCherry purified protein. Single z-stack movies were acquired in a Zeiss LSM 780 in Airyscan mode, with a Zeiss C-Apochromat 40×/1.2 W Korr FCS M27, 5× zoom and 0.5 s time interval.

Dynein inhibition

For dynein inhibition, p150 was injected at a final concentration of 7 mg ml⁻¹ in late-one-cell-stage embryos (after cell expansion). Under these conditions, p150 had a nearly immediate effect: injection in the late one-cell stage ensured that the first mitosis had already occurred and the effect of p150 would be detected at the two-cell stage. P150 resulted in shorter metaphase spindles but no impact on metaphase chromosome congression. During anaphase A, chromosomes separated without significant differences from control embryos, but anaphase B was clearly affected. After anaphase, cells failed cytokinesis, resulting in a four-cell-stage embryo with only two cells, each cell with two spindles. The analysis of p150-injected embryos was restricted to two-cell-stage embryos that showed the phenotype described above. To analyse flows after dynein inhibition, p150 was co-injected with EB3–mCherry (to analyse the spindle phenotype) and WGA-640 on Tg(Ef1α:MLS-GFP) transgenic embryos.

Inhibition of actin polymerization

Total inhibition of actin polymerization led to the loss of embryo shape and was incompatible with cell viability. We achieved inhibition of cytoplasmic actin without affecting dramatically embryo shape (and cortical actin) with a 3 min incubation of embryos in Danieau 0.3% solution with latrunculin B (428020; Sigma-Aldrich) at a final concentration of 5 μM.

Cdk1 inhibition

For Cdk1 inhibition, embryos were incubated for 5 min in Danieau 0.3% with dinaciclib²⁰ at 200 μm (CAY-14707; Cayman Chemical). With this concentration, entry in mitosis was delayed, confirming the inhibition of Cdk1 activity.

Light-induced microtubule depolymerization

For light-induced microtubule depolymerization, we used SBTub3-P (ref. ²⁴), a soluble, non-reversible and 405-nm-activatable version of the initially published photostatin⁴⁴. SBTub3-P was injected at one-cell-stage embryos. The injection and mounting of embryos was performed under red-light conditions to avoid premature activation of the drug. Drug activation was performed at the metaphase/anaphase onset with three to four pulses of 405 nm laser light through the entire z-stack of imaging (25 μm). This ensured normal metaphase spindle assembly and proper chromosome congression and did not substantially affect chromosome segregation during anaphase A. SBTub3-P was combined with live imaging of Tg(bactin2:HsENSCONSIN17-282-3xEGFP) to analyse the effect on microtubules during metaphase and anaphase or live imaging of Tg(Ef1α:MLS-GFP) transgenic embryos and co-injected with α-tubulin-ATTO-565 and WGA-640 to analyse cytoplasmic flows after induced microtubule depolymerization.

Endogenous lipid droplets

Endogenous lipid droplets were labelled with Nile Red. Embryos were incubated for 5 min in Danieau 0.3% solution with Nile Red (72485, Sigma-Aldrich) at a final concentration of 10 μM.

Exogenous lipid droplets

Exogenous lipid droplets were assembled according to published protocols⁴⁵. In brief, in a bovine serum albumin (BSA)-precoated glass flask, 1 ml of DSPE-PEG(2000)-biotin (1,2-distearoyl-sn-glycero-3-phosphoethanolamine-N- (biotinyl(polyethylene glycol)-2000) was added at a final concentration of 2 mM. The PEG solution was sonicated for 1 min. With a BSA-precoated tip, 70 μl of FC70 oil was added to the PEG solution. After vigorous shaking and pipetting, the emulsion was formed. The resulting mix is composed of droplets of various sizes and can be maintained over several weeks at 4 °C. Droplets were fluorescently labelled with Cy3-streptavidin (PA43001, Cytiva Amersham) and injected in the cell of one-cell-stage embryos.

Cells without DNA

Cells without DNA were identified initially in Tg(Ef1α:MLS-GFP) transgenic embryos with α-tubulin-ATTO-565 and H1-488 co-injected. One cell showed no H1-488 labelling and the absence of a metaphase spindle, while the neighbouring cells had both DNA and a metaphase spindle. The absence of DNA or a mitotic spindle was not a consequence of lack of labelling as all other cells in the same embryo had labelling. In Xenopus egg extracts, the Ran-GTP gradient was shown to be essential for metaphase spindle assembly⁴⁶. The gradient is generated on mitotic chromosomes⁴⁷. Thus, in zebrafish embryos, in the cells without DNA the gradient could be impaired, explaining the lack of mitotic spindle. The fate of these cells and how they arise in the embryo are not understood.

Yolk aspiration

Yolk aspiration was performed during the first 30 min after fertilization, before the ooplasm streaming contribution to the final cell size. A microinjection needle (1 mm glass capillary; TW100F-3, Word Precision Instruments) was assembled on a PicoNozzle kit (version 2, 5430-All, World Precision Instruments), which is, in turn, connected to a 10 ml syringe. This creates enough vacuum pressure to aspirate yolk material from several embryos. Embryos were displayed in a line, in a similar fashion as for embryo injection, and aspiration was performed until an approximately 50% reduction of yolk size was observed. For live imaging experiments, embryos from different genetic backgrounds were aspirated and subsequently injected with different markers. The injection droplet was much reduced (~5 μm) compared with the final size of the embryos and, therefore, did not change the effect of the aspiration.

Embryo patterning

A 3D-printed template (adapted from Donoughe et al.⁴⁸) was used to pattern rectangularly shaped boxes (900 × 250 × 1,500 μm³) on 2% agarose. Two-cell-stage wild-type embryos were always maintained in Danieau 0.3% solution while being inserted into the confined spaces (with the help of forceps) and were covered with a coverslip to maintain the shape during imaging.

Immunofluorescence

Whole embryos with chorions were fixed with 4% paraformaldehyde for at least 5 h at room temperature. After fixation, embryos were washed 2× with phosphate-buffered saline (PBS)–Tween 0.05% and manually dechorionated. Permeabilization was done with PBS–Triton 0.3% for 10 min, followed by blocking with 5% BSA in PBS–Tween 0.05% for 1 h, overnight primary antibody incubation, three 5 min washes with PBS–Tween 0.05%, and overnight incubation with secondary antibody. Antibody incubations were done in 5% BSA in PBS–Tween 0.05%. After secondary antibody, three 10 min washes with PBS–Tween 0.05% were performed. DAPI was added in the last wash. The primary antibody was rabbit anti-pH3-s10 D2C8 (1:200; 3377; Cell Signalling (lot 7)), and the secondary antibody was anti-rabbit Alexa-488 (1:200).

Data analysis

Systematic analysis of scaling of NER

NER position and timing was systematically analysed in Tg(Ef1α: H2B-mCherry) transgenic embryos from 4- to 512-cell stage co-injected with NLS-GFP and WGA-640. The distance between the two sets of DNA that separate during anaphase was defined as the chromosome distance. The chromosome distance when WGA or NLS are recruited to chromosomes was defined as WGA or NER, respectively. NER timings were defined as seconds after anaphase onset (t = 0 s, anaphase onset). One- and two-cell stages were not included in this analysis owing to the low number of replicates.

Quantification of tubulin speckles

Before tracking, movies were processed with background subtraction and a one-pixel Gaussian blur filter. Speckles were automatically tracked with TrackMate plug-in on Fiji⁴⁹ using the DoG (difference of Gaussian) detector and a 0.5 μm blob diameter, and spots were linked with a simple LAP tracker. Only tracks with a minimum of three spots were considered. Statistical analysis was done with GraphPad Prism.

Quantification of EB3 microtubule growth

Single stack movies were acquired every 0.5 s. Before tracking, movies were processed with background subtraction and a one-pixel Gaussian blur filter. The microtubule growth was tracked with TrackMate plug-in on Fiji⁵⁰. In brief, a LoG detector with a blob diameter of 1 µm was used. Tracks were linked with the simple LAP (linear assignment problem) tracker. Obtained tracks were filtered for a minimum of three spots and with a linearity threshold of 0.6. Statistical analysis was done with GraphPad Prism.

Quantification of astral microtubule growth

Astral microtubule growth velocity was obtained from measurements of aster diameter, perpendicular to the axis of chromosome separation, for a time interval between 40 s before to 40 s after anaphase onset. DCX–GFP transgenic embryos were used.

Quantification of aster anisotropy

Aster anisotropy was calculated as the ratio of the distance between the aster centre and the spindle midzone (d_a) and the radius of the aster (r_a) (see also the Supplementary Information). Measurements were performed at 100 s after anaphase onset (corresponding to the start of anaphase B) using DCX–GFP transgenic embryos.

Analysis of pH3-s10 profiles

Analysis of the phosphorylation gradient was performed on movies of embryos injected with pH3-s10 fluorescently labelled with 488, 561 or 640, depending on the combination with other markers. Sum projections were used in the analysis. Fluorescence intensity was measured in a circular ROI of constant size in the chromosome region. This fluorescent signal was background subtracted, measured with the same ROI outside the chromosome region. Intensity values were further processed in MATLAB with custom-made codes; namely, single intensity profiles were normalized to the maximum value within the first 80 s after anaphase onset. A single exponential function (,left(right.C(T)={C}_{0}{{mathrm{e}}}^{-T/tau })) was fitted to the average curve of all profiles from all cell stages as a function of time. From the fit we extracted the decay time (τ) and the dephosphorylation rate (1/τ). pH3-s10 live imaging was performed in Tg(Ef1α: H2B-mCherry) transgenic embryos to use as the internal control H2B–mCherry. Profile intensities were obtained using the same ROIs and analysis pipeline as for pH3-s10. Indeed, H2B–mCherry profiles show a minor fluorescence intensity decay that we attribute to chromosome decondensation, compared with the decay of pH3-s10.

MSD of mitochondria tracks

Individual mitochondria were manually tracked with the Manual Tracking plug-in from Fiji. Only tracks that showed the initial fast episode followed by a slow episode were considered for tracking and analysis. Full tracks were exported into Excel and manually segmented into fast versus slow episodes. MSD (mean square displacement) analysis was performed using the MSD analyser⁵⁰, a MATLAB-based code. Both episodes on each track were fitted with a parabolic function (({{rm{MSD}}}left(tright)=4{Dt}+{v}^{2}{t}^{2})), from which the velocity and diffusion were obtained.

PIV analysis

Flows were visualized with mitochondria present in the cytoplasm, labelled with a MLS–GFP transgenic line¹⁶. From the movies obtained, a series of 20 frames, corresponding to the timings from anaphase onset until 200 s after anaphase onset, were selected for PIV analysis. A subset of three to four z-stacks, in the plane of chromosome separation, was used for maximum projection. Flows were analysed using PIVlab, version 2.55 (ref. ⁵¹), a MATLAB-based software. In brief, images were preprocessed with a contrast-limited adaptive histogram equalisation (CLAHE) algorithm using a 64-pixel window. Because the mitotic spindle and chromosome region excluded mitochondria localization, a mask was drawn manually to exclude these regions from the PIV analysis. A ROI including the total size of the cell was defined to apply PIV analysis only to the cell of interest. PIV analysis was performed with a fast fourier transform cross correlation algorithm using three multipass windows of 128 × 128, 64 × 64 and 32 × 32 pixels. A velocity threshold was defined for each obtained flow field to exclude outliers and a minimal smoothening was applied. The final flow field was averaged from 100–200 s, when the flows are more prominent. To obtain flow velocity near the chromosomes, WGA-640 signal was used as a marker for chromosome position. A mask was defined manually in the place where WGA labels the chromosomes. Flow velocity ‘near the chromosomes’ corresponds to the velocity values 9 μm above and below the chromosome mask.

Statistics and reproducibility

No statistical methods were used to predetermine sample sizes, but our sample sizes are similar to those reported in previous publications⁴. The experiments were not randomized and the investigators were not blinded to allocation during experiments and analysis of the data. Normality of the samples was determined with a D’Agostino and Pearson test. Statistical analysis for two-sample comparison, with normal or non-normal distribution, was performed with a t-test or Mann–Whitney test, respectively. For multiple-group comparison, a parametric one-way analysis of variance or a non-parametric analysis of variance (Kruskal–Wallis) was used, for samples with normal or non-normal distribution, respectively. All pairwise multiple comparisons were subsequently analysed using either Tukey’s (parametric) or Dunn’s (non-parametric) tests. All statistical analysis was performed with GraphPad Prism V5 (GraphPad Software). Micrographs are representative of a set of at least two independent experimental rounds.

Kymographs

Kymographs were generated using a previously published custom-written MATLAB code⁸.

Software

For image acquisition, Slidebook 6 (3i) and Zeiss Zen were used. Numerical simulations were performed with custom-written code on Julia version 1.10.5 (version of Julia packages is detailed on GitHub). Image processing and analysis were performed with FiJi (2.9) and MATLAB 2021b (The MathWorks). For statistics and data analysis, Microsoft Excel for Mac 2023, Prism 10 was used. Figures were assembled using Affinity Photo v1 and Affinity Publisher v1.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.