To investigate chromosome positioning during early mitosis, we quantified how widely chromosomes were scattered in space. To do this, we acquired three-dimensional images of fluorescently labeled DNA (SiR-DNA) in live mitotic human U2OS cells. Subsequently, we obtained the convex hull of the chromosome distribution. The convex hull in two and three dimension (2D and 3D) is the minimal polygon and polyhedron, respectively, that wrap multiple objects (chromosomes in this case). This is schematically shown in Figure 1A, using a shrinking ‘rubber band’ and ‘balloon’ for intuitive demonstration. We determined the volume of this minimal polyhedron or shrunk balloon that wrapped chromosomes, and defined it as the chromosome scattering volume (CSV), which represents how widely chromosomes scatter in 3D space.

Figure 1 with 5 supplements see all

Download asset Open asset

Figure 1—source data 1

Data at individual time points in individual cells.

https://doi.org/10.7554/eLife.46902.008

Download elife-46902-fig1-data1-v3.xlsx

To synchronize U2OS cells in early mitosis, we used a previously described U2OS cell line in which the CDK1 gene was replaced, by genome engineering, with cdk1-as, whose gene product Cdk1-as can be specifically inhibited by an ATP analogue 1NM-PP1 (Rata et al., 2018). The U2OS cdk1-as cells were arrested in G2 with 1NM-PP1, and subsequently released by 1NM-PP1 washout to synchronously undergo mitosis. NEBD was identified by the release of the nuclear-localizing fluorescent reporter (GFP-LacI-NLS) into the cytoplasm (Figure 1—figure supplement 1). Using these methods, we quantified the CSV in synchronized U2OS cdk1-as cells as they progressed from NEBD (defined as t = 0) into prometaphase. We found that CSV was prominently reduced within the first 8 min following NEBD, and reduction continued more slowly over the following ~10 min (Figure 1B). CSV reduction was also observed in asynchronous, wild-type U2OS cells (Figure 1—figure supplement 2), indicating that it was not an artifact caused by cdk1-as-mediated cell cycle synchronization. Consequently, we used U2OS cdk1-as cells for synchronous entry into mitosis for the rest of the experiments presented in this work, unless otherwise stated.

It is possible that the CSV reduction observed following NEBD was an outcome of the interaction between chromosomes and spindle MTs. To test this possibility, we quantified CSV in cells treated with nocodazole. After nocodazole treatment, MTs were almost completely depleted (Figure 1—figure supplement 3). In control cells, chromosomes moved inward after NEBD, and subsequently aligned on the metaphase plate (Figure 1C, top-row images; Video 1). In cells lacking visible MTs, chromosomes also moved inward after NEBD, but then remained in a spherical formation (Figure 1C, bottom-row images; Video 1; Figure 1—figure supplement 4). The CSV was reduced with very similar kinetics in both the presence and absence of MTs (Figure 1C, graphs). CSV may be slightly smaller at 4 min following NEBD in control cells; if so, this could be due to mild compression of the nuclear envelope (NE) remnants caused by the rapid MT-dependent inward movement of centrosomes (Figure 1C, image at +4 min in control). Consequently, we conclude that the overall CSV reduction observed following NEBD is not an outcome of chromosome interaction with spindle MTs.

Video 1

Download asset

Alternatively, chromosome condensation, which occurs in early mitosis, may cause the CSV reduction after NEBD. To test this possibility, we quantified the total volume of chromosome mass (chromosome mass volume [CMV]; Figure 1—figure supplement 5) after NEBD. Unlike the CSV, the CMV did not considerably change following NEBD. Thus, we conclude that the reduction in CSV was not due to chromosome condensation, which likely occurs mainly before NEBD (Mora-Bermúdez et al., 2007).

What mechanisms, then, might promote CSV reduction after NEBD? The CSV reduction could be explained if chromosomes were surrounded by a physical barrier that contracts following NEBD. Consistent with this model, we found that in early mitosis an actin network accumulated on the NE, peaking in intensity around NEBD and persisting after it (Figure 2A; Video 2). The volume inside this actin network was rapidly reduced within 6 min following NEBD (Figure 2—figure supplement 1), suggesting the contraction of the network. During this time, the chromosomes were surrounded by, and the outermost chromosomes were found immediately adjacent to, the actin network (Figure 2—figure supplement 2; Video 3). We used cdk1-as in these experiments to synchronize cells, however the actin network was also observed in wild-type U2OS cells (Figure 1—figure supplement 2, yellow arrowheads), indicating that it is not an artifact due to cdk1-as regulation.

Figure 2 with 2 supplements see all

Download asset Open asset

Figure 2—source data 1

Data at individual time points in individual cells.

https://doi.org/10.7554/eLife.46902.013

Download elife-46902-fig2-data1-v3.xlsx

Video 2

Download asset

Video 3

Download asset

We investigated in more detail the localization of the actin network on the NE or its remnants, following NEBD. Using Airyscan super-resolution microscopy, we compared the localization of the actin network and the lamin B1 network; the latter is found on the nucleoplasmic surface of the NE (Ungricht and Kutay, 2017) and remains substantially intact soon after NEBD (Georgatos et al., 1997). The actin network was separated outwardly from lamin B1 by 150–200 nm (Figure 2B), suggesting that it may localize on the cytoplasmic surface of the NE (or its remnants) following NEBD. A candidate receptor for the actin network on the NE is the LINC (linker of nucleoskeleton and cytoskeleton) complex, one end of which is anchored in the NE while the other end extends to the cytoplasm and interacts with actin (Luxton and Starr, 2014). To test whether the NE-association of the actin network during early mitosis was LINC complex-dependent, we visualized the actin network in cells expressing a previously described dominant negative construct of the LINC complex (LINC-DN) or its control – SR-KASH and KASH∆L, respectively, reported in (Luxton et al., 2010). As expected, both constructs localized to the NE (Figure 2C, yellow arrowheads). Crucially, the fluorescence intensity of the actin network was significantly diminished in cells expressing LINC-DN, compared with cells expressing its control (Figure 2C, magenta arrowheads, graph). Therefore, the actin network localizes to the cytoplasmic surface of the NE and its remnants, following NEBD, in a LINC complex-dependent manner.

If the actin network were a physical barrier, confining chromosomes, and its contraction were to lead to the CSV reduction after NEBD, LINC complex disruption should also alleviate CSV reduction. We tested this hypothesis by measuring the CSV in cells expressing LINC-DN and its control (Figure 2D; Video 4) and found that the CSV reduction was incomplete after NEBD with the LINC-DN expression, compared with the control expression. Thus, it is likely that the actin network on NE remnants confines chromosomes and that its contraction leads to CSV reduction.

Video 4

Download asset

We next addressed how the actin network contracts on the NE remnants to reduce the CSV after NEBD. We first tested whether actin depolymerization causes the CSV reduction. An actin depolymerization inhibitor jasplakinolide had no effect on CSV reduction after NEBD (Figure 3—figure supplement 1), although it successfully inhibited subsequent cytokinesis as previously reported (Murthy and Wadsworth, 2005). Thus it is unlikely that actin depolymerization promotes actin network contraction or CSV reduction.

The actin-dependent motor protein myosin II is a major mediator of acto-myosin contractility in non-muscle cells (Svitkina, 2018). Thus, we next tested the possible involvement of myosin II in the actin network contraction and CSV reduction. We found that myosin II co-localized with the actin network on the NE (Figure 3—figure supplement 2, yellow arrowheads). Given this, we next inhibited myosin II activity using para-nitroblebbistatin (pnBB), a photo-stable version of the myosin II inhibitor blebbistatin, which is suitable for live-cell imaging (Képiró et al., 2014). Notably, pnBB treatment alleviated both the CSV reduction (Figure 3A; Video 5) and the actin network contraction (Figure 3—figure supplement 3A,B) following NEBD, which suggests that these processes are dependent on the myosin II activity.

Figure 3 with 6 supplements see all

Download asset Open asset

Figure 3—source data 1

Data at individual time points in individual cells.

https://doi.org/10.7554/eLife.46902.024

Download elife-46902-fig3-data1-v3.xlsx

Video 5

Download asset

On the other hand, the pnBB treatment did not affect CMV (see Figure 1—figure supplement 5) after NEBD (Figure 3—figure supplement 4). This suggests that, while myosin II activity facilitates CSV reduction, this reduction is not due to enhanced chromosome condensation. Moreover, pnBB treatment did not significantly change the distance between spindle poles soon after NEBD (Figure 3—figure supplement 3A,C). As far as we observed, pnBB-treated cells did not seem to show considerably altered shapes of the spindle (Figure 3—figure supplements 3 and 5), compared with control (and LINC-DN-expressing) cells (Figure 3—figure supplement 6), following NEBD. Thus, it seems unlikely that the alleviated CSV reduction due to pnBB treatment was the outcome of an abnormal mitotic spindle. Nonetheless, pnBB treatment often caused the actin network to extend beyond the spindle poles at the initiation of spindle formation (Figure 3—figure supplement 5, pnBB, 5–20 min).

Furthermore, the pnBB treatment alleviated the CSV reduction in the presence of nocodazole (Figure 3B; Video 6), indicating that myosin II promotes actin network contraction and CSV reduction after NEBD in a spindle MT-independent manner. Double treatment with pnBB and nocodazole seemed to further alleviate the CSV reduction during 0–6 min, compared with pnBB treatment alone (compare Figure 3A and B). Thus, a MT-dependent mechanism, leading to mild NE remnant compression by centrosomes (see Figure 1C), and a myosin II-dependent mechanism may work in concert to reduce CSV during this period.

Video 6

Download asset

A straightforward interpretation of these results is that myosin II within the actin network causes its contraction and the CSV reduction after NEBD. To obtain more evidence for this model, we performed light-induced local inhibition of myosin II activity in mitotic cells. We used azido-blebbistatin (azBB), which covalently binds the heavy chain of myosin II and inhibits its ATPase activity when exposed to infra-red light (Képiró et al., 2012). A low concentration of azBB was used to ensure that myosin II was inhibited only at the region exposed to infra-red light. We exposed a half of the nucleus to infra-red light (Figure 4A) and compared the CSV over time in the exposed and non-exposed halves of the nucleus. In the exposed half of the nucleus (Figure 4B,C; blue), the CSV reduction was alleviated, compared to the non-exposed half (red). In the cells not treated with azBB (control), there was no significant difference in CSV between exposed (Figure 4B,C; green) and non-exposed (orange) halves of nuclei. These results support a model in which myosin II activity within the actin network (hereafter referred to as the ‘acto-myosin’ network) on NE remnants causes its contraction and CSV reduction after NEBD.

Figure 4

Download asset Open asset

Figure 4—source data 1

Data at individual time points in individual cells.

https://doi.org/10.7554/eLife.46902.028

Download elife-46902-fig4-data1-v3.xlsx

Following NEBD, spindle MTs interact with chromosomes while CSV is reduced by contraction of the acto-myosin network associated with NE remnants (Figure 1B). The acto-myosin network contraction might facilitate correct interaction between chromosomes and spindle MTs, leading to subsequent chromosome congression to the middle of the mitotic spindle. To test this, we measured the period between NEBD and completion of chromosome congression in cells expressing either LINC-DN or its control. LINC-DN expression led to a considerable delay in chromosome congression and anaphase onset, compared with the control (Figure 5A–C; Video 7), which was not associated with significant defects in spindle positioning or length (Figure 5—figure supplement 1). In cells expressing LINC-DN, chromosomes often remained behind a spindle pole and outside of the pole-to-pole region after NEBD (back-of-spindle chromosomes) (Figure 5A bottom and D, magenta arrowheads), which is consistent with the extension of NE remnants beyond spindle poles when contraction of the acto-myosin network was suppressed (Figure 3—figure supplement 5). In a subgroup of cells expressing LINC-DN where back-of-spindle chromosomes still remained at 20 min after NEBD in cells, there was a larger delay in congression (Figure 5—figure supplement 2). Moreover, LINC-DN expression led to more frequent abnormal segregation of chromosomes during anaphase (Figure 5E).

Figure 5 with 4 supplements see all

Download asset Open asset

Figure 5—source data 1

Data in individual cells.

https://doi.org/10.7554/eLife.46902.034

Download elife-46902-fig5-data1-v3.xlsx

Video 7

Download asset

The above results with LINC-DN expression were reproduced with pnBB treatment (which inhibits myosin II) (Figure 5—figure supplement 3). Collectively, these results suggest that the CSV reduction due to contraction of the acto-myosin network diminishes back-of-spindle chromosomes and facilitates correct chromosome interaction with spindle MTs. These effects likely lead to timely chromosome congression and anaphase onset as well as high-fidelity chromosome segregation.

The above studies used human U2OS (near triploid) cells. We also investigated the acto-myosin network on the NE in other human cell lines. RPE cells (diploid) showed accumulation of actin on the NE in early mitosis (Figure 5—figure supplement 4, top). By contrast, HeLa cells (highly aneuploid) did not show such NE-associated actin accumulation (Figure 5—figure supplement 4, bottom). An interesting possibility is that a lack of the acto-myosin network is correlated with aneuploidy.

To synchronize U2OS cells in early mitosis with a chemical genetic approach, their CDK1 genes were replaced with cdk1-as construct (Rata et al., 2018), as previously done in DT40 cells (Hochegger et al., 2007). The U2OS cdk1-as cells, HeLa cells (ATCC CCL-2) and original U2OS cells (ATCC HTB-96) were cultured in DMEM (Gibco 41965–039) medium supplemented with 10% FBS (Gibco 10500) and 1% 10,000 U/ml Pen/strep (Gibco 15140–122). hTERT RPE (ATCC CRL-4000) were maintained in DMEM-F12 medium (Gibco 11330–057) supplemented with 10% FBS (Gibco 10500) and 1% 10,000 U/ml Pen/strep (Gibco 15140–122). To detach cells from culture flasks or dishes, they were treated with 0.05% Trypsin (Gibco 25300–054). All cell lines used in this study were verified by Eurofins authentication service, using STR profiling. There was no contamination of mycoplasma in cell cultures in tests using Mycoalert mycoplasma detection kits (Lonza LT07-118).

Plasmid DNA constructs used in this study were: GFP-LacI-NLS (Hori et al., 2013; pT2699), mCherry-Lifeact (a gift from Michael Davidson, Addgene #54491; pT3188), GFP-Tubulin (a gift from Jason Swedlow lab; pT2932), LINC-DN (mRFP1-SR-KASH; Luxton et al., 2010; pT3138) and LINC-DN Control (mRFP1-KASH-ΔL; Luxton et al., 2010; pT3139); H2B-Cerulean (a gift from Sean Megason lab; pT2698), and MHC-GFP (CMV-GFP-NMHC II-A, a gift from Robert Adelstein, AddGene #11347; pT3153). For transfection of plasmids, reactions were prepared in 135 µl OptiMem medium plus Glutamax (Gibco 51985–026) with 3 µl DNA FuGene HD Transfection Reagent (Promega E2311) for every 1 µg of plasmid DNA.

Small molecule inhibitors were used with the following concentrations: 1NM-PP1 1 µM (Merck Millipore 529581), RO-3306 10 µM (Merck Millipore 217699), nocodazole 3.3 µM (Sigma M1404), paranitroblebbistatin (pnBB) 50 µM (motorpharmacology), azidoblebbistatin (azBB) 5 µM (motorpharmacology), jasplakinolide 1 µM (Merck J4580) and MK-1775 500 nM (Selleck Chemicals S1525). A Wee1 inhibitor MK-1775 was used to facilitate entry into mitosis by bypassing G2/M checkpoint (Hirai et al., 2009) in our experiments with nocodazole (Figures 1C and 3B) and with LINC-DN plus GFP-LacI-NLS (Figure 2D). Use of MK-1775 alone did not significantly change kinetics of CSV reduction – compare Figure 2D, red (with MK-1775) and Figure 3A, red (without MK-1775).

Live-cell images and immunofluorescence images were acquired using a Deltavision Elite widefield microscopy system (GE Healthcare) with 100x and 60x objective lenses (Olympus, NA 1.4) and cameras Cascade 1K EMCCD (Roper Scientific) and CoolsnapHQ2 (Photometrics). For live-cell imaging, the microscope chamber was maintained at 37°C with 5% carbon dioxide. For photoactivation of azBB, we used Zeiss 710 confocal microscopy system with a Coherent Chameleon multiphoton laser attachment and with 63x objective lens (NA 1.4). Super-resolution images were acquired using a Zeiss Airyscan 880 system with 63x objective lens (NA 1.4). Widefield Images were deconvolved using SoftWorx and images were analyzed using Fiji, OMERO, Imaris and Volocity software.

U2OS cdk1-as cells were synchronized and prepared for live-cell imaging as follows: cells were plated out at 50–60% confluency onto glass-bottomed imaging dishes (FluoroDish; World Precision Instruments, FD35-100). The following evening they were transfected with the relevant plasmids after replacement of the media with media lacking pen/strep. The transfection was allowed to continue overnight before the transfection reagents were washed out with further media without pen/strep. Cells were then arrested in the evening by addition of 1NM-PP1 (in the case of U2OS cdk1-as cells) or RO-3306 (in the case of HeLa or RPE cells), alongside SiR-DNA to stain chromatin where required. At this stage, media was also replaced with Fluorobrite medium (Gibco A18967-01) supplemented with 10% FBS (Gibco 10500), 2 mM L-Glutamine (Gibco 200 mM, 25030–025), 25 mM HEPES (1 m Lonza BE17-737E) and 1 mM Na-Pyruvate (100 mM, Lonza BE13-115E). The following day cells were released from arrest by washing in 2 ml Fluorobrite medium (with supplements) 12 times, immediately followed by image acquisition. pnBB, azBB and MK-1775 were added after the final wash, prior to imaging. Nocodazole was added 1 hr before release, and added again after the final wash, prior to imaging. GFP-Tubulin construct was stably transfected in cells shown in Figure 2A, Figure 2—figure supplement 2, Figure 5A, Figure 3—figure supplements 3, 5 and 6, and Figure 5—figure supplement 3. Images were acquired every 2 min except for a) Figure 2A, Figure 2—figure supplement 2 and Figure 3—figure supplement 5 where images were acquired every minute, and b) Figure 1—figure supplement 2 where images were acquired every 5 min. During imaging, 10–12 z-sections were acquired with 1.6 µm z-intervals, except for a) Figure 2A, Figure 1—figure supplement 3, Figure 3—figure supplement 2 and Figure 5—figure supplement 4, where 10–12 z-sections were acquired 0.8 µm z-intervals and b) Figure 2—figure supplement 2 and Figure 3—figure supplement 5, where 5 z-sections were acquired with 1.6 µm z-intervals. For live-cell imaging of asynchronous U2OS cells, they were treated in the same way but were not arrested overnight before imaging. When a cell drifted during live-cell imaging, the frame of time sequence images in figures was adjusted so that a chromosome mass was still positioned at the center of the frame in each image. The timing of NEBD was determined by (a) the efflux of GFP-LacI-NLS signals from the nucleus into the cytoplasm, (b) the influx of GFP tubulin signals from the cytoplasm into the nucleus, or (c) the influx of GFP-NMHCII-A signals from the cytoplasm into the nucleus.

The night before fixation, media was replaced with Fluorobrite medium with supplements and the cells arrested with 1NM-PP1 if required. On the following day, cells were released by washing in 2 mL Fluorobrite medium 12 times. Progression into mitosis was confirmed under a phase contrast light microscope. Then cells were fixed when they reached an appropriate stage, based on observation of chromosome condensation. To fix cells for immuno-staining, they were rinsed with pre-warmed PBS and incubated for 10 min with 37°C 4% methanol-free formaldehyde at 37°C. They were then rinsed three times with PBS and permeabilized for 10 min in PBS with 0.5% Triton X-100. Following this they were blocked with 5% BSA in PBS for an hour at 37 °C. While blocking, primary antibodies were prepared in 5% BSA in PBS to the relevant dilution. Following blocking, cells were incubated rocking with the primary antibody mixture for at least 60 mins at room temperature (or 4°C if incubated overnight). Following primary incubation, cells were washed rocking in PBS for 5 min, three times. Meanwhile, secondary antibodies plus phalloidin/Hoechst 33342 were prepared to the relevant dilution in 5% BSA in PBS. Cells were then incubated rocking with secondary antibodies for at least 60 min at room temperature (or 4 °C if incubated overnight). Following secondary incubation, cells were washed rocking in PBS for 5 min, three times.

Primary antibodies were used for immuno-staining with the following dilution: anti-Tubulin 1:1000 (Cell Signalling Technology YL1/2), anti-LaminB1 1:1000 (Abcam ab16048) and anti-Actin 1:1000 (Thermo-Fisher PA1-183). Secondary antibodies were used with the following dilution: A488 goat anti-Rabbit 1:1000 (Thermo-Fisher A-11034), A488 donkey anti-Rat 1:1000 (Thermo-Fisher A-21208), and A594 goat anti-Rabbit 1:1000 (Thermo-Fisher A-11037). Dyes were used for cell staining with the following dilution: Phalloidin DyLight 650 1:1000 (Invitrogen 13454309), Hoechst 33342 1:2000 (Sigma-Aldrich 14533), SiR-DNA 100 nm (tebu-bio SC007).

U2OS cdk1-as cells expressing GFP-LacI-NLS were prepared for live-cell imaging as described above. SiR-DNA was supplemented when 1NM-PP1 was added to arrest cells. Azidoblebbistatin (azBB) was added when cells were released from the arrest (by washing off 1NM-PP1). To identify cells in prophase, the start of chromosome condensation was monitored with SiR-DNA signal. A scan area was zoomed in so that it covered roughly half of the nucleus in prophase. A z-stack (three sections, 2 µm apart) of the region was then rapidly scanned four times with 633 nm and 488 nm lasers to observe chromosomes and GFP-LacI-NLS, respectively, alongside 860 nm multi-photon laser exposure to photoactivate the covalent binding of azBB to Myosin II. The region was scanned in three rounds in this way at 1-min interval. The scan area was then extended so that image of the entire nucleus could be acquired. The whole nucleus was then imaged for SiR-DNA and GFP-LacI-NLS in a z-stack of 20 µm with 2 µm z-interval, every 2 min. The timing of NEBD was judged based on the efflux of GFP-LacI-NLS from the nucleus into the cytoplasm. The nuclear region exposed to 860 nm light could be discriminated based on partial photo-bleaching of SiR-DNA signals.

U2OS cdk1-as cells expressing GFP-LacI-NLS were imaged every 2 min in the presence of SiR-DNA. Z-stacks of 10-sections were acquired with 2 µm interval. The CSV was measured as follows: The surface of chromosomes was identified using Imaris (3D surface object tool). Obviously non-chromosome objects were excluded in this process. On each z-section, the surface contours of multiple chromosomes were connected into one single-stroke contour (Figure 1B, green lines imposed on images), using Imaris, in such a way that the connection doesn’t affect 2D convex hull. These contours were used to generate a continuous 3D surface contour, which was required to compute a 3D convex hull. Based on this continuous 3D surface contour, a 3D convex hull was generated using Imaris (convex hull generation tool). The volume of this 3D convex hull was defined as the CSV. To compare reduction of CSV between two conditions, we used two-way ANOVA in which data at all the time points shown in relevant graphs were included in analyses.

U2OS cdk1-as cells expressing mCherry-Lifeact and GFP-tubulin were imaged every 2 min. Z-stacks of 10-sections were acquired at 2 µm intervals. A 3D surface object was generated using Imaris by creating contours following the actin network at each z-section. The volume of the surface object was taken as the volume inside the actin network. Meanwhile, to measure the intensity of actin network, a two pixel wide line was drawn along the actin network on a single z-section using Fiji; the shape of the nucleus as seen by the exclusion of GFP-tubulin background was used as a guide when the actin signals were ambiguous. The mean gray value of the pixels within this line was taken as a measurement. As a background reading, a measurement was taken from a similar two-pixel wide line drawn in the cytoplasmic region surrounding the nucleus. To generate a final reading, the measurement from the cytoplasm was subtracted from the measurement from the actin network.

U2OS cdk1-as cells imaged every 2 min in the presence of SiR-DNA. Z-stacks of 10 sections were acquired with 2 µm interval. A 3D surface covering SiR-DNA signal was automatically generated with Imaris as follows; the mean value in the Cy5 channel was set as the threshold, and the Background Subtraction (local contrast) option was chosen. The volume of the surface object was defined as the chromosome mass volume (CMV).

U2OS cdk1-as cells expressing GFP-tubulin and H2B-Cerulean were imaged every 2 min. Z-stacks of 10-sections were acquired with 2 µm interval. Timing of complete congression was estimated as the time all chromosomes aligned on the spindle equator. Timing of anaphase onset was estimated as the time chromosomes began segregation to opposite poles.

All experiments were repeated at least twice and similar results were obtained. Statistical analyses were carried out using Prism software (Graphpad). Methods of statistics are stated in each relevant figure legend. All p-values were two-tailed.

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

We thank Tanaka lab members, A Ciulli, A Testa, J Januschke and M Gierlinski for discussion; G Ball for help with image analysis; L Clayton for editing the manuscript;, T Fukagawa, J Swedlow, S Megason, R Adelstein, M Davidson, and RY Tsien for reagents; S Swift, and P Appleton for microscope maintenance. This work was supported by the Wellcome Trust (096535/Z/11/Z, 097945/Z/11/Z, 208401/Z/17/Z), Cancer Research UK (C28206/A114499) and Medical Research Council (MR/K015869/1). TUT is a Wellcome Trust Principal Research Fellow. The authors declare no competing financial interests.

© 2019, Booth et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.