A mitotic chromatin phase transition prevents perforation by microtubules

Dividing eukaryotic cells package extremely long chromosomal DNA molecules into discrete bodies to enable microtubule-mediated transport of one genome copy to each of the newly forming daughter cells1–3. Assembly of mitotic chromosomes involves DNA looping by condensin4–8 and chromatin compaction by global histone deacetylation9–13. Although condensin confers mechanical resistance to spindle pulling forces14–16, it is not known how histone deacetylation affects material properties and, as a consequence, segregation mechanics of mitotic chromosomes. Here we show how global histone deacetylation at the onset of mitosis induces a chromatin-intrinsic phase transition that endows chromosomes with the physical characteristics necessary for their precise movement during cell division. Deacetylation-mediated compaction of chromatin forms a structure dense in negative charge and allows mitotic chromosomes to resist perforation by microtubules as they are pushed to the metaphase plate. By contrast, hyperacetylated mitotic chromosomes lack a defined surface boundary, are frequently perforated by microtubules and are prone to missegregation. Our study highlights the different contributions of DNA loop formation and chromatin phase separation to genome segregation in dividing cells.

Dividing eukaryotic cells package extremely long chromosomal DNA molecules into discrete bodies to enable microtubule-mediated transport of one genome copy to each of the newly forming daughter cells [1][2][3] . Assembly of mitotic chromosomes involves DNA looping by condensin [4][5][6][7][8] and chromatin compaction by global histone deacetylation [9][10][11][12][13] . Although condensin confers mechanical resistance to spindle pulling forces [14][15][16] , it is not known how histone deacetylation affects material properties and, as a consequence, segregation mechanics of mitotic chromosomes. Here we show how global histone deacetylation at the onset of mitosis induces a chromatin-intrinsic phase transition that endows chromosomes with the physical characteristics necessary for their precise movement during cell division. Deacetylation-mediated compaction of chromatin forms a structure dense in negative charge and allows mitotic chromosomes to resist perforation by microtubules as they are pushed to the metaphase plate. By contrast, hyperacetylated mitotic chromosomes lack a defined surface boundary, are frequently perforated by microtubules and are prone to missegregation. Our study highlights the different contributions of DNA loop formation and chromatin phase separation to genome segregation in dividing cells.
The material properties of individual cell components have a key role in the dynamic self-organization of cellular structures. In mitotic vertebrate cells, chromosomes must acquire material properties that enable microtubules to move them, first to the spindle centre during prometaphase and then to the spindle poles during anaphase 1,17 . Microtubules attach to and pull on chromosomes at specialized kinetochore regions 18 , whereas microtubules contacting chromosome arms generate polar ejection forces that push chromosomes away from the spindle poles [19][20][21][22] , generating a complex system with high tension 17,23 . Condensins cross-link mitotic chromosomes to confer the mechanical stability required to withstand the tension generated at kinetochores [14][15][16] , but it remains unclear how chromosomes acquire material properties that enable them to resist, and therefore move in response to, polar ejection forces. These material properties must provide chromosome arms with sufficient resistance to prevent penetration by polymerizing microtubule tips, as microtubules growing through the chromatin fibre loops in mitotic chromosomes would result in entanglements that impair segregation.

Mitotic chromatin excludes microtubules
To investigate how chromosomes resist microtubules acting on their arms, we first studied in human tissue culture cells (HeLa cells) the effect of condensin depletion on the morphology and movement of mitotic chromosomes. To deplete condensin, we modified all endogenous alleles of its essential structural maintenance of chromosomes 4 (SMC4) subunit with a C-terminal auxin-inducible degron (mAID) and HaloTag for visualization and added the auxin analogue 5-PhIAA at 2.5 h before mitotic entry to induce efficient degradation (Extended Data Fig. 1a-c). We visualized the spindle using silicon-rhodamine (SiR)-tubulin and stained DNA with Hoechst 33342 to determine the position of chromosomes. In condensin-depleted mitotic cells, we observed an unstructured mass of compact chromatin forming a plate between the spindle poles ( Fig. 1a,b and Extended Data Fig. 1d,e). Immunofluorescence staining of the kinetochore marker centromere protein A (CENP-A) and the spindle pole component pericentrin further showed that many kinetochores were detached from the bulk mass of chromatin and displaced towards the spindle poles, whereas, in control cells, all of the kinetochores were closely linked with chromosomes at the metaphase plate (Extended Data Fig. 1f,g). These observations confirm that condensin is required for chromosomes to resist tension generated at kinetochores [14][15][16] and further show that bulk chromatin is positioned at the spindle centre by a mechanism that is independent of tightly associated kinetochores.
The chromatin of condensin-depleted cells might be moved by polar ejection forces. To study how chromosomes are moved by the mitotic spindle, we imaged mitotic entry of HeLa cells stably expressing mCherry-tagged core histone 2B (H2B-mCherry) and eGFP-tagged Article CENP-A, and visualized microtubules by SiR-tubulin. As it is difficult to distinguish the effect of poleward and anti-poleward forces in a bipolar spindle configuration, we induced a monopolar spindle geometry by inhibiting kinesin-5 using S-trityl-l-cysteine (STLC). When condensin-expressing control cells entered mitosis, the chromosomes first moved towards the spindle pole shortly after nuclear disassembly and then arranged in a rosette with the kinetochores facing towards the pole and the chromosome arms facing away from the pole, such that the region surrounding the spindle pole remained free of chromosomes (Extended Data Fig. 2a,d and Supplementary Video 1). This arrangement is consistent with a balance between pole-directed microtubule pulling at the kinetochores and polar ejection forces pushing on the chromosome arms [19][20][21][22][23] . When condensin-depleted cells entered mitosis, chromatin formed a compact mass that moved away from the spindle pole, whereas the kinetochores approached the spindle pole, resulting in the detachment of a large fraction of kinetochores from the bulk mass of chromatin (Extended Data Fig. 2b,d,e and Supplementary Video 2). Thus, condensin-depleted chromatin remains responsive to polar ejection forces, whereas it is not stiff enough to resist the tension generated at the kinetochores.
we treated condensin-depleted cells with 5 µM of the pan histone deacetylase inhibitor trichostatin A (TSA) 2.5 h before mitotic entry to induce broad hyperacetylation of histone lysine residues (Extended Data Fig. 3a-d), whereby the short duration of the treatment did not induce apoptosis or DNA double-stranded breaks (Extended Data Fig. 3e-h). The hyperacetylated chromatin of condensin-depleted mitotic cells did not compact or enrich at the spindle centre but instead diffusely distributed throughout the cytoplasm, such that spindle pole regions were no longer clear of chromatin (Fig. 1a,b and Extended Data Fig. 1d-g). Treatment with 500 nM TSA also effectively suppressed chromatin compaction and localization to the spindle centre, whereas treatment with 500 nM TSA followed by 8 h removal of TSA resulted in normal chromosome morphology in mitosis, validating the specificity and reversibility of the phenotypes (Extended Data Fig. 4a-d).
A complementary approach to induce histone hyperacetylation by overexpressing the histone acetyltransferase p300 also resulted in chromatin decompaction phenotypes similar to those induced by TSA (Extended Data Fig. 5). Thus, histone deacetylation is important for chromatin compaction and positioning at the spindle centre.
To investigate more specifically how hyperacetylation affects the response of chromosomes to polar ejection forces, we imaged mitotic entry of live condensin-depleted cells treated with TSA, using STLC to induce a monopolar spindle geometry. The chromatin of these cells remained diffuse and completely decompacted, while the spindle aster assembled and moved into the decondensed chromatin regions. Kinetochores then moved towards the spindle pole, but the bulk mass of chromatin was not displaced towards the cell periphery such that the regions surrounding the spindle poles did not clear from chromatin, and the kinetochores remained embedded in the chromatin (Extended Data Fig. 2c-e and Supplementary Video 3). Thus, deacetylation has an important role in the response of chromatin to polar ejection forces.
The diffuse distribution of chromatin throughout the cytoplasm resulting from TSA treatment in condensin-depleted cells is in stark contrast to relatively mild mitotic chromosome decompaction phenotypes previously observed in condensin-expressing wild-type cells 9,10 . We hypothesized that the moderate level of TSA-induced decompaction in wild-type cells might be due to condensin-mediated linkages that counteract the dispersion of chromatin fibres. To investigate this, we analysed how TSA affects chromatin density in the presence or absence of condensin. In condensin-depleted mitotic cells, TSA reduced chromatin density to 29% compared with cells that were not treated with TSA, whereas, in the presence of condensin, TSA reduced chromatin density only to 53% and chromosomes remained visible as thread-like structures (Fig. 1a,c and Extended Data Fig. 4a-d). Condensin depletion alone did not reduce mitotic chromatin density (Fig. 1a,c), consistent with previous observations 14,25,26 . Thus, histone deacetylation is necessary and sufficient for complete compaction of mitotic chromatin even in the absence of condensin. By contrast, condensin is neither necessary nor sufficient for complete chromatin compaction during mitosis, yet it can concentrate chromatin to some extent even when histones are hyperacetylated.
We next investigated whether chromatin compaction through histone deacetylation might be necessary to prevent microtubules from penetrating chromosomes. To investigate how acetylation affects the access of microtubules to chromosomes, we performed electron tomography of mitotic HeLa cells. Chromosomes of unperturbed cells appeared as homogeneously compacted bodies with a sharp surface boundary, and 3D segmentation showed that they were almost never penetrated by microtubules (Fig. 1d,e and Extended Data Fig. 6a). By contrast, chromosomes of TSA-treated cells appeared to be less compact, particularly towards the periphery, and microtubules grew extensively through the chromatin (Fig. 1d-f and Extended Data Fig. 6b-d). Thus, active histone deacetylases are required to keep microtubules out of chromosome bodies, providing a basis for resistance towards polar ejection forces.
Microtubule perforation into mitotic chromosomes is expected to cause entanglements between chromatin fibre loops and spindle microtubules that impair chromosome segregation. To investigate how TSA-induced hyperacetylation alone affects chromosome segregation, we recorded time-lapse videos of HeLa cells expressing H2B-mCherry. TSA severely delayed chromosome congression and initiation of anaphase and caused a high incidence of lagging chromosomes (Extended Data Fig. 6e-h). Active histone deacetylases are therefore essential for faithful chromosome segregation.

A mitotic chromatin phase transition
To elucidate the mechanism that underlies microtubule exclusion from mitotic chromosomes, we investigated how acetylation affects the material properties of chromatin. Recent research has demonstrated that phase-separated biomolecular condensates can form highly dense structures that exert and resist forces 27 . Moreover, we found that purified nucleosome arrays condense into liquid droplets in physiological salt solutions, and these condensates dissolve after acetylation 13 , supporting the idea that mitotic chromatin might form an immiscible phase. However, endogenous chromatin contains thousands of different proteins 28 and is subject to various post-translational modifications besides acetylation 12,24 , with unknown effects on phase separation. Under which conditions endogenous chromatin might undergo a phase transition, how such chromatin phase transition might affect the material properties of chromosomes and what might be the functional relevance are unclear.
To test the hypothesis that mitotic chromosomes are highly complex biomolecular condensates, we reasoned that fragmenting mitotic chromatin using a nuclease might relieve constraints imposed by the very long length of chromosomes to unveil the underlying phase transition. To investigate this, we developed a live-cell chromatin fragmentation assay based on microinjection of the restriction enzyme AluI. Shortly after microinjection of AluI into mitotic cells, chromosomes lost their elongated shape, forming round condensates that fused to one another, consistent with a liquid-like state ( Fig. 2a and Supplementary Video 4). Notably, AluI injection did not decrease chromatin density (Fig. 2b). As chromosome fragmentation is expected to induce cell death in the long run, we imaged chromatin only a few minutes after AluI injection and validated by the early apoptosis marker polarity sensitive indicator of viability and apoptosis (pSIVA) that within this short time frame cells do not enter apoptosis (Extended Data Fig. 7a,b). Overall, these experiments show that the integrity of the chromatin fibre is not required for full compaction of the mitotic chromatin, consistent with a phase separation mechanism of compaction.
To assess the mobility of chromatin, we performed a fluorescence recovery after photobleaching analysis of H2B-mCherry. Native mitotic chromosomes recovered very little H2B-mCherry fluorescence after photobleaching, consistent with constrained mobility within a large polymer network. However, after AluI digestion, H2B-mCherry recovered rapidly and completely from photobleaching (Fig. 2c,d), consistent with a liquid state. Imaging Halo-tagged SMC4 further showed that condensin did not form axial structures inside the chromatin condensates and instead evenly distributed throughout the cell, validating efficient chromosome fragmentation by AluI (Extended Data Fig. 7c,d). Thus, mitotic chromatin is insoluble in the cytoplasm and, when the long-range constraints of the fibre network are eliminated, the short-range dynamics manifest in liquid-like behaviour.
To test whether the formation of an immiscible chromatin phase is suppressed by acetylation, we treated cells with TSA before mitotic entry and then injected AluI. This resulted in homogeneously dispersed chromatin fragments with almost no local condensates (Fig. 2e,f and Supplementary Video 5; the few remaining chromatin foci might represent constitutive heterochromatin that is known to be refractory to TSA-induced hyperacetylation 29 ); a similar phenotype was observed Article in cells overexpressing p300 to induce histone hyperacetylation (Extended Data Fig. 7e,f). By contrast, depletion of SMC4 had no effect on chromatin droplet formation after AluI-mediated fragmentation (Extended Data Fig. 7g,h). Thus, deacetylation is a major factor in establishing an immiscible chromatin phase in mitotic cells, while condensin is not required.
We next investigated whether chromatin fragments generated in interphase nuclei undergo a solubility phase transition after progression to mitosis. To monitor cell cycle stages, we used a fluorescence resonance energy transfer (FRET) biosensor for a key mitotic kinase, aurora B. As chromosome fragmentation blocks mitotic entry owing to DNA damage signalling, we applied chemical inhibitors to induce an interphase-to-mitosis transition. We first synchronized cells to G2 using the cyclin dependent kinase 1 (CDK1) inhibitor RO3306 and then induced a mitosis-like state by removing RO3306 for CDK1 activation and simultaneously inhibiting counteracting protein phosphatase 2 (PP2A) and protein phosphatase 1 (PP1) using okadaic acid. Mitotic entry was demonstrated by the aurora B FRET biosensor signal (Extended Data Fig. 8a,b and Supplementary Video 6). Injection of AluI into G2 cell nuclei resulted in homogeneously distributed chromatin, consistent with a soluble state; furthermore, after induction of mitosis, chromatin fragments formed spherical condensates that were as dense as intact chromosomes (Extended Data Fig. 8c,d and Supplementary Video 7), whereas control cells in which RO3306 was not replaced by Okadaic acid maintained homogeneously dissolved chromatin fragments (Extended Data Fig. 8e,f). These observations support a model in which a global reduction in solubility drives chromatin compaction at the interphase-to-mitosis transition.
To determine whether the loss of chromatin solubility after mitotic entry depends on deacetylation, we inhibited histone deacetylases using TSA before injecting AluI into G2-synchronized cells. As in control cells, chromatin fragments were homogeneously distributed throughout the nucleus in G2, but mitotic induction did not lead to the formation of condensed foci (Extended Data Fig. 8g  The bars indicate the mean. Significance was tested using two-tailed Mann-Whitney U-tests (P = 9.305 × 10 −10 (TSA); P = 0.476 (AluI)). Biological replicates: n = 3 (a,b,g-i); n = 2 (c-f). Scale bars, 5 µm (a, e and g, main images), 1 µm (a, e and g, insets) and 3 µm (c).
the effect of histone acetylation on chromatin solubility more specifically, we used synthetic nucleosome arrays 13 as probes. We generated fluorescently labelled arrays of 12 unmodified naive nucleosomes as well as similar arrays that were labelled with a distinct fluorophore and acetylated in vitro with recombinant p300 acetyltransferase. In vitro, the unmodified nucleosome arrays form liquid condensates under physiological salt concentrations, in contrast to the acetylated nucleosome arrays 13 (Extended Data Fig. 8i,j). After co-injection of these nucleosome arrays into live mitotic cells, unmodified arrays almost completely partitioned into the mitotic chromatin phase, whereas acetylated nucleosome arrays predominantly dissolved in the cytoplasm (Extended Data Fig. 8k,l). Thus, acetylation-sensitive interchromatin interactions alone are sufficient to recruit chromatin into mitotic chromosomes, supporting a model in which histone acetylation is a direct regulator of chromatin solubility in the cytoplasm.
To further characterize the boundary between the chromatin and cytoplasm, we studied a component of the mitotic chromosome periphery-the protein Ki-67 [30][31][32] . Ki-67 has an N-terminal region that is excluded from mitotic chromatin and a C-terminal region that is attracted to mitotic chromatin 31 . According to our chromatin phase-separation model, the targeting of Ki-67 to the chromosome surface should arise from its amphiphilic attraction to the phase boundary between the chromatin and cytoplasm and therefore be independent of higher-order chromatin fibre folding but sensitive to chromatin solubilization. To test this hypothesis, we studied the localization of Ki-67 in live cells, finding that Ki-67 still enriched at a sharp boundary around chromatin droplets after AluI injection or around chromatin of cells depleted of condensin, whereas TSA substantially reduced confinement of Ki-67 to the chromatin surface ( Fig. 2e-g and Extended Data Fig. 8m,n). We also tested whether Ki-67 is required for the formation of an insoluble chromatin phase. After injection of AluI into MKI67-knockout (encoding Ki-67) cells 31 , we observed spherical chromatin condensates similar to those in wild-type cells (Extended Data Fig. 8o,p). Thus, Ki-67 targets to the surface of mitotic chromosomes through its amphiphilic attraction to chromatin and cytoplasmic phases, whereas it is not required for the formation of an immiscible mitotic chromatin phase.
Overall, our data show that deacetylation during mitotic entry induces global chromatin phase separation. This chromatin phase separation mediates full compaction of mitotic chromatin, independently of the integrity of the chromatin fibre, condensin-mediated DNA looping 4-8 or potential higher-order chromatin fibre coils [33][34][35] . Chromosomes have been described as hydrogels, in which a flexible chromatin fibre is cross-linked by condensin and expanded throughout its volume by an aqueous liquid component [36][37][38][39][40] . By elucidating acetylation as a key regulator of chromatin solubility, our study provides a molecular explanation of how such chromatin hydrogels collapse into compact bodies with a sharp boundary in mitosis.

Macromolecular exclusion from chromatin
To investigate the mechanism underlying microtubule exclusion from mitotic chromosomes, we analysed how soluble tubulin partitions relative to mitotic chromatin. We microinjected fluorescently labelled tubulin into live mitotic cells and applied nocodazole to suppress microtubule polymerization (Fig. 3a). Soluble tubulin was much less concentrated inside mitotic chromosomes compared with in the surrounding cytoplasm (Fig. 3a,b). By contrast, soluble tubulin was not excluded from hyperacetylated chromosomes in TSA-treated cells (Fig. 3a,b). Thus, the immiscible chromatin compartment formed by deacetylated mitotic chromatin excludes soluble tubulin.
To determine whether the exclusion of tubulin is due to a limiting pore size in chromosomes, we expressed DsRed, a fluorescent protein that forms tetramers slightly larger than tubulin dimers 41 . In contrast to tubulin, DsRed distributed evenly across chromatin and cytoplasm ( Fig. 3c,d). Thus, macromolecules in the size range of tubulin are not generally excluded from mitotic chromatin.
The selective exclusion of tubulin but not DsRed from mitotic chromatin suggests that specific molecular features control access. Tubulin is highly negatively charged at physiological pH (7.2), in contrast to near-neutrally charged DsRed, raising the possibility that macromolecular access is governed by electrostatic interactions. A high concentration of an overall negative electrical charge of chromatin 42 in an immiscible condensate might therefore repel negatively charged Article cytoplasmic proteins. To investigate how charge affects macromolecular access to mitotic chromatin, we fused charged polypeptides to the N terminus of DsRed. DsRed fused to a negatively charged polypeptide (overall predicted charge on the tetramer, −7e) was excluded from mitotic chromatin, whereas DsRed fused to a positively charged polypeptide (overall charge on the tetramer, +9e) concentrated inside chromatin (Fig. 3c,d). Similarly, we found that negatively charged monomeric enhanced green fluorescent protein (meGFP) or negatively charged 4 kDa dextrans were also excluded from mitotic chromosomes, whereas a positively charged surface mutant scGFP (super-charged GFP, +7e) 43 or positively charged 4 kDa dextrans concentrated in the mitotic chromosomes (Extended Data Fig. 9a-d). Two negatively charged microtubule-regulating proteins, p80-katanin and the microtubule plus-tip protein end binding 3 (EB3), were also excluded from the mitotic chromatin, whereby TSA treatment resulted in a less efficient exclusion from the mitotic chromatin (Extended Data Fig. 9e-h). To investigate how molecular size affects the partitioning relative to mitotic chromatin, we injected negatively charged dextrans of 20 or 70 kDa molecular mass. Both dextrans were excluded from chromatin, and the exclusion was more efficient at greater molecular mass; TSA treatment reduced the exclusion of both dextrans from mitotic chromosomes (Extended Data Fig. 9i-l). Thus, electrical charge and molecular size are key determinants of macromolecular access to mitotic chromatin. The exclusion of negatively charged macromolecules from mitotic chromosomes might be mediated by nucleosome fibres alone, or it might involve other chromosome-associated factors. To investigate how tubulin interacts with intrinsic chromatin condensates formed by reconstituted nucleosome arrays, we reconstituted droplets of recombinant nucleosome arrays in vitro 13 and then added rhodamine-labelled tubulin in the presence of nocodazole. Soluble tubulin was indeed efficiently excluded from nucleosome array condensates (Fig. 3e,f). Consistent with this observation, negatively charged meGFP or dextrans were also excluded from nucleosome array condensates, whereas a positively charged scGFP mutant or positively charged dextrans concentrated in nucleosome array condensates (Extended Data Fig. 9m-p). The exclusion of negatively charged macromolecules, including tubulin, is therefore an intrinsic property of condensed nucleosome fibres.
The efficient exclusion of free tubulin suggests that weak affinity interactions in liquid chromatin condensates might be sufficient to limit microtubule polymerization. To investigate how microtubules interact with reconstituted chromatin droplets in vitro, we added purified tubulin and then induced microtubule polymerization by increasing the temperature. Microtubules formed a dense microtubule network, yet they almost never grew into chromatin condensates (Fig. 3e,f). Thus, chromatin-intrinsic material properties impose a highly impermeable barrier to microtubule polymerization independently of condensin or other chromosome-associated factors.

Microtubules push liquid chromatin
Polymerizing microtubules exert substantial pushing forces after contact with stiff surfaces 44 . As microtubules did not grow into condensates of purified chromatin fragments in vitro, we wondered whether the surface tension of liquified endogenous chromatin is strong enough to allow microtubule-based pushing. If this were the case, then droplets of digested chromosomes should be pushed away by polar ejection forces of growing astral microtubules. To test this hypothesis, we injected AluI into mitotic cells treated with nocodazole and then washed out the nocodazole to induce microtubule polymerization, while applying STLC to induce a monopolar astral spindle geometry. Time-lapse imaging showed that, initially, all chromatin resided in a single droplet at the cell centre but, soon after nocodazole washout, the chromatin split into several droplets that moved away from growing microtubule asters (Extended Data Fig. 10a,b). Thus, liquid chromatin condensates can be pushed by polymerizing astral microtubules. Astral microtubules might directly push on chromatin droplets by polymerizing against the chromatin phase boundary or they might couple to chromatin through chromokinesins [19][20][21][22][23][45][46][47][48] . To investigate how the spindle moves chromatin droplets, we used RNA interference (RNAi) to co-deplete two chromokinesins that are major contributors to the polar ejection force, Kid and kinesin family member 4A (Kif4a) 22,23,45,47,48 . After AluI-injection and nocodazole washout in the presence of STLC, chromosome droplets moved to the chromosome periphery as efficiently as in the control cells (Extended Data Fig. 10c-e). Thus, spindle asters can move liquid chromatin independently of the chromokinesins Kid and Kif4a, potentially by polymerizing microtubules pushing against the chromatin phase boundary.
Polar ejection forces pushing on chromosome arms are counteracted by pole-directed pulling at centromeres. To investigate how liquified chromatin responds to pulling forces at the kinetochores, we injected AluI into cells expressing eGFP-CENP-A and H2B-mCherry in the presence of nocodazole. We then removed nocodazole to induce monopolar spindle assembly in the presence of STLC. The bulk mass of chromatin, visualized through H2B-mCherry, rapidly moved towards the cell periphery, whereas several much smaller chromatin condensates enriched in eGFP-CENP-A remained close to the spindle monopole ( Fig. 4a,b and Supplementary Video 9). Thus, bulk chromatin is pushed away from the spindle poles, while centromeres are transported towards the spindle pole. When the continuous connection between centromeres and the remaining chromosome is lost, bulk chromatin and centromeric chromatin physically separate according to the locally dominating forces.

Conclusions
Here we show that a substantial global reduction in chromatin solubility caused by deacetylation during mitotic entry converts chromosomes into phase-separated bodies rather than loose bottlebrush structures with chromatin fibre loops extending into the cytoplasm (Fig. 4c). The immiscible mitotic chromatin forms a surface that provides resistance to microtubule perforation while allowing local chromatin fibre sliding internally, as required for continuous dynamic loop formation by condensin 6,14,49 . In parallel, condensin-mediated linkages establish a hydrogel that withstands tension generated at kinetochores [14][15][16] . Jointly, these molecular activities shape discrete chromosome bodies with a defined surface despite continuous internal remodelling of the chromatin fibre.
Our results show that mitotic cells contain three principal domains with distinct microtubule polymerization propensity: the centrosome matrix, which concentrates soluble tubulin to promote nucleation at the poles 50 ; the cytoplasm, which is highly permissive for microtubule growth and amplification; and a chromatin compartment that is refractory towards polymerization. Thus, our study provides a unified view of how chromatin looping by condensin and compaction through a phase transition driven by acetylation-sensitive nucleosome interactions contribute to the material properties and mechanical functions of mitotic chromosomes. It will be interesting to investigate how chromatin adapts its material properties to other physiological processes that involve compaction, such as apoptosis.

Online content
Any methods, additional references, Nature Research reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41586-022-05027-y. Publisher's note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.

Cell lines and cell culture
All of the cell lines used in this study have been regularly tested negative for mycoplasm contamination. All cell lines in this study were derived from a HeLa 'Kyoto' cell line that was previously described in ref. 51  Live-cell imaging was performed in DMEM containing 10% (v/v) FBS (Gibco, 10270-106, 2078432), 1% (v/v) penicillin-streptomycin (Sigma-Aldrich) and 1% (v/v) GlutaMAX (Gibco; 35050038) but omitting phenol red and riboflavin to reduce autofluorescence (imaging medium) 51 . Cells were grown in 10 cm or 15 cm Cellstar (Greiner) dishes, in 75 cm 2 or 175 cm 2 Nunc EasYFlask cell culture flasks (Thermo Fisher Scientific) or in 96-well, 48-well, 12-well or 6-well Nunclon Delta Surface multiwell plates (Thermo Fisher Scientific). For live-cell imaging, cells were cultivated on Nunc LabTek II chambered cover glass (Thermo Fisher Scientific), on µ-Slide 8-well covered coverslips (Ibidi), in µ-Dish 35 mm high imaging dishes with a polymer or glass bottom (Ibidi).

Generation of stable fluorescent reporter and genetically engineered cell lines
Cell lines stably expressing fluorescently labelled marker proteins were generated by random plasmid integration (for the transfection conditions, see below) or by a lentiviral vector system pseudotyped with a mouse ecotropic envelope that is rodent-restricted (RIEP receptor system). Construction of the RIEP receptor parental cell lines and the subsequent generation of stable cell lines that express fluorescent marker proteins was performed as described previously 54 .
HeLa genome editing was performed using a CRISPR-Cas9-mediated integration approach, using the chimeric Cas9-human Geminin fusion (Cas9-hGem) modification for enhanced editing efficiency 55 . Single-guide RNA (sgRNA) was cloned into pX330-U6-Chimeric_ BB-CBh-hSpCas9-hGem-P2A-mCherry*aBpil (gift from S. Ameres). For tagging of endogenous SMC4 genes with an auxin-inducible degron tag 56 , a repair template for targeting the C terminus of the protein was designed, with homology flanks of 800 and 901 bp length for the 5′ and 3′ flanks, respectively. The repair template containing the coding sequence for a C-terminal mAID-HaloTag, including mutations of the protospacer adjacent motif, was synthesized as a gBlock (IDT) and cloned into plasmid pCR2.1 for amplification. sgRNA/Cas9 and homology repair template plasmids were co-transfected using the Neon transfection system (Thermo Fisher Scientific). Transfection was performed according to the manufacturer's guidelines (for HeLa cells: pulse voltage, 1.005 V; pulse width, 35 ms; pulse number, 2), using the 100 µl tip cuvette, with 10 µg homology repair template and 10 µg gRNA/Cas9-hGEM for 1 × 10 6 cells. Then, 9 days after electroporation, cell pools were stained with Oregon Green HaloTag (Promega, G2801) ligand and single clones were isolated using fluorescence-activated cell sorting (FACS) and sorted into 96-well plates. Genotyping was performed as in a previous study 54 .
Integration of the OsTIR1(F74G) ligase into the adeno-associated virus integration site (AAVS1, designated safe harbour) locus for establishment of AID2 auxin-inducible degradation of SMC4 was achieved using a sgRNA and homology repair template strategy described in ref. 57 . Expression of the E3-ubiquitin ligase OsTIR1(F74G) enables degradation of degron-tagged proteins by labelling them for proteasomal degradation in the presence of a small-molecule auxin analogue. sgRNA was cloned into pX330-U6-Chimeric_BB-CBh-hSpCas9-hGem-P2A-mCherr y*aBpil. For integration of the OsTIR1(F74G), a homology template containing an OsTIR1(F74G)-SnapTAG-MycA1-NLS and homology flanks of 804 and 837 bp length for the 5′ and 3′ flanks, respectively, was generated. Co-transfection of sgRNA/Cas9 and homology repair template plasmids was performed as described above. Then, 8 days after electroporation, transfected cells were selected with 6 µg ml −1 blasticidin. Single clones were isolated from the stable pool by single-cell dilution into 96-well plates. To identify clones that depleted SMC4-mAID-Halo, clones were treated for 90 min with 1 µM 5-Ph-IAA24, stained with Oregon Green HaloTag (Promega, G2801) ligand and then analysed by flow cytometry using the iQue Screener Plus system.

Transfection of plasmids and small interfering RNA
For expression of fluorescently labelled markers, the respective genes were cloned into bicistronic vectors containing an IRES or a 2A 'self-cleaving' peptide sequence with antibiotic resistance genes that enable the protein of interest and the resistance gene to be expressed from the same transcript. For transient transfection or transfection for subsequent selection and colony picking, plasmids were transfected using X-tremeGENE 9 DNA transfection reagent (Roche, 6365787001) according to the manufacturer's protocol (1 µg plasmid, 4.5 µl X-tremeGENE 9 in 100 µl serum-free OptiMEM) or PEI transfection reagent (1 mg ml −1 stock, Polyscience 24765-2, 4 µg of transfection reagent per 1 µg of plasmid). For stable expression, plasmids were transfected using PEI, 10 µg plasmid for one ~70% confluent 15 cm dish and incubated for 48 h before antibiotic selection.
Small interfering RNAs (siRNAs) were delivered with lipofectamine RNAiMax (Invitrogen) according to the manufacturer's instructions. hKid (Kif22) was targeted using 16 nM custom silencer select siRNA (sense strand CAAGCUCACUCGCCUAUUGTT, Thermo Fisher Scientific, including a 3′ overhang TT dinucleotide for increased efficiency). Kif4A was targeted using 16 nM custom silencer select siRNA (sense strand GCAAGAUCCUGAAAGAGAUTT, Thermo Fisher Scientific, including a 3′ overhang TT dinucleotide for increased efficiency). Custom silencer select siXWNeg (sense strand UACGACCGGUCUAUCGUAGTT, Thermo Fisher Scientific, including a 3′ overhanging TT dinucleotide for increased efficiency) was used as a non-targeting siRNA control. Cells were imaged 30 h after siRNA transfection. Co-transfection of KIF22 and KIF4A siRNAs was performed using 16 nM final concentrations each. Both hKid and KIF4A siRNAs were described in ref. 22 .

Inhibitors and small molecules
To degrade Smc4-mAID-Halotag, cells were incubated in 1 µM 5-phenylindole-3-acetic acid (5-PhIAA) (Bio Academia, 30-003) for 2.5-3 h. To induce histone hyperacetylation, cells were incubated with 5 µM TSA 58 (Sigma-Aldrich, T8552). To arrest cells in prometaphase with monopolar spindle configuration, cells were incubated in 5 µM STLC 59 (Enzo Life Sciences, ALX-105-011-M500) for 2-3 h. To arrest cells in prometaphase, cells were incubated for 2-3 h in nocodazole (200 ng ml −1 for live-cell imaging only) or 100 ng ml −1 for microinjection and/or subsequent washout; Sigma-Aldrich, M1404). Nocodazole washout was performed by washing cells five times with prewarmed imaging medium on the microscope. To arrest cells in metaphase for microinjection, cells were incubated with 12 µM proTAME (R&D systems, I-440-01M) for 2 h. To synchronize cells to the G2-M boundary, cells were first synchronized with a double thymidine block followed by a single RO3306 60 block. One day after seeding, cells were transferred into medium containing 2 mM thymidine (Sigma-Aldrich, T1895). Then, 16 h later, cells were washed three times with prewarmed medium and released for 8 h. The thymidine block was repeated once. Then, 6 h after the second thymidine release cells were transferred into medium containing 8 µM RO3306 (Sigma-Aldrich, SML0569) for 4 h. After G2 arrest, RO3306 was removed, and the cells were washed three times with imaging medium containing 1.5 µM okadaic acid (LC laboratories, O-5857). To induce apoptosis as a positive control for apoptotic index measurements, cells were treated with 5 µM anisomycin (Sigma-Aldrich, A9789) for 3 h. To induce DNA double-stranded breaks as a positive control for DNA damage measurements, cells were treated with 50 ng ml −1 neocazinostatin (Sigma-Aldrich, N9162) for 3 h.

Preparation of 384-well microscopy plates and coverslips for in vitro assays
For microtubule/nucleosome array droplet experiments (Fig. 3e), thin layer cover glass sandwiches were constructed from passivated 24 × 60 mm Menzel coverglass (VWR). To clean coverslips, they were vertically stacked into a Coplin jar (Canfortlab, LG084). Coplin jars were then sonicated in acetone (Sigma-Aldrich) for 15 min, in 100% ethanol (VWR) for 15 min and then washed 10 times with ≥18 MΩ MonoQ H 2 O. All of the sonication steps were performed using an ultrasonic cleaning bath (Branson, 2800 Series Ultrasonic Cleaner, M2800). In a separate Erlenmeyer flask, 31.5 ml 30% aqueous hydrogen peroxide solution (Sigma-Aldrich, 31642) was added to 58.5 ml concentrated sulfuric acid (Sigma-Aldrich, 258105) (piranha acid). The flask was gently swirled until bubbling and heating occurred, then the whole contents of the flask was added to the coverslips in the Coplin jars, ensuring that all coverslip surfaces were covered. The jar was transferred into a 95 °C water bath and heated for 1 h. Afterwards, the piranha solution was discarded, and the cover glasses were washed once with ≥18 MΩ MonoQ H 2 O and etched with 0.1 M KOH for 5 min. The cover glasses were transferred to a fresh, dry Coplin jar and dried to completion in a 65 °C benchtop oven, and afterwards left to cool to room temperature. In a separate Erlenmeyer flask, 4 ml dichlorodimethylsilane (DCDMS) (Sigma-Aldrich, 440272, anhydrous) was injected into 80 ml heptane (Sigma-Aldrich, 246654, anhydrous). The contents of the flask were immediately transferred to the jar containing the cover glasses, and the jar was incubated for 1 h at room temperature. Next, silane was decanted, and the cover glasses were sonicated in chloroform (Sigma-Aldrich) for 5 min, washed in chloroform once, sonicated in chloroform again for 5 min and sonicated twice in ≥18 MΩ MonoQ H 2 O. Finally, the cover glasses were washed once more in chloroform, air dried and stored in a sealed container in a dry and dark, dust-free space (for up to 6 months).

Expression and purification of GFP proteins
pET-based constructs for the expression and purification of meGFP(−7e) and scGFP(+7e) were a gift from D. Liu 43 . An overnight culture of Rosetta 2 (pLysS) Escherichia coli (Novagen) transformed with pET_−7GFP or pET_+7GFP plasmids encoding GFP with a theoretical peptide charge of −7e or +7e, respectively, were grown on an agar plate by replating a single transformant on LB supplemented with 100 ng µl −1 of ampicillin at 37 °C. The bacterial lawn was suspended in LB supplemented with the 100 ng µl −1 of ampicillin and grown at 37 °C to an optical density at 600 nm of 0.4, cooled over 1 h to 18 °C, and recombinant protein expression was induced by addition of IPTG to 0.5 mM for 18 h at 18 °C. The cells were collected by centrifugation, resuspended in NiNTA lysis buffer (50 mM HEPES·NaOH, pH 7, 150 mM NaCl, 10% (w/v) glycerol, 15 mM imidazole, 5 mM β-mercaptoethanol, 1 mM benzamidine, 100 µM leupeptin, 100 µM antipain, 1 µM pepstatin), the cellular suspension was flash frozen in liquid N 2 and stored at −80 °C.

Nucleosome array in vitro experiments
Generation, fluorophore labelling, assembly and quality control of nucleosome arrays has been described previously 13 . Here 12X601 nucleosome arrays carrying no label or an Alexa Fluor 488 or Alexa Fluor 594 label were used. The covalently fluorescently labelled nucleosome arrays used for microinjection were labelled at an 8% fluorophore density (8 in 100 histone2B proteins labelled with a fluorophore) and dialysed against TE (10 mM Tris-HCl (Sigma-Aldrich), pH 7.4 (Sigma-Aldrich), 1 mM EDTA (Sigma-Aldrich), pH 8.0, 1 mM DTT (Roche, 10708984001)) to remove glycerol.
To visualize soluble tubulin partitioning relative to chromatin droplets in vitro, 20 µl of chromatin droplet suspension was transferred to a passivated well of a 384-well microscopy plate. Chromatin droplets were allowed to sediment for 15 min. Next, 5 µl of soluble TRITC-labelled tubulin (Cytoskeleton, TL590M) in phase-separation buffer containing 500 ng ml −1 nocodazole was added for a final tubulin concentration of 5 µM. Soluble tubulin was equilibrated for 10 min before images were recorded.
To visualize GFP surface charge variant partitioning relative to chromatin droplets in vitro, 20 µl of chromatin droplet suspension was transferred to a passivated well of a 384-well microscopy plate. Chromatin droplets were allowed to sediment for 15 min, after which 5 µl of GFP solution in phase-separation buffer was added to a final concentration of 5 µM. GFP proteins were equilibrated for 10 min before images were recorded.
To visualize chemically modified dextran partitioning into chromatin droplets in vitro, 20 µl of chromatin droplet suspension was transferred to a passivated well of a 384-well microscopy plate. Chromatin droplets were allowed to sediment for 15 min, after which 5 µl dextran solution in phase-separation buffer was added to a final concentration of 500 µg ml −1 . The dextran used was a fluorescein (FITC)-labelled 4.4 kDa dextran fraction (negative overall charge conferred to dextran by fluorophore charge) (Sigma-Aldrich, FD4) or a FITC-labelled 4.4 kDa dextran fraction modified with diethylaminoethyl (DEAE) groups conferring an overall positive charge (TdB, DD4).
All of the images of macromolecule partitioning relative to chromatin droplets in vitro ( Fig. 3e and Extended data Fig. 9m,o) were recorded on the incubated stage of a customized Zeiss LSM980 microscope using a ×40 1.4 NA oil-immersion DIC plan-apochromat objectives (Zeiss).

Nucleosome array acetylation in vitro
To generate acetylated nucleosome arrays for microinjection, recombinant p300 histone acetyl transferase domain was generated (VBCF Protein technologies Facility) using plasmid pETduet+p300 HAT according to a purification strategy described previously 13 . To induce histone acetylation of nucleosome arrays, arrays with 8% Alexa Fluor 488 or Alexa Fluor 594 label density at a concentration of 3.85 µM were incubated with 6.12 µM p300-HAT (enzyme stock 61.2 µM in gel-filtration buffer (20 mM Tris·HCl, pH 8.0, 150 mM NaCl, 10% (w/v) glycerol, 1 mM DTT)) in the presence of 750 µM Acetyl-CoA (Sigma-Aldrich, A2056) in gel-filtration buffer at room temperature for 2 h with occasional agitation. The acetylation was next stopped by the addition of A-485 (Tocris, 6387) to a final concentration of 9 µM and the reaction mixture was stored in the dark. To 10 µl of quenched acetylation reaction, 10 µl dilution buffer containing 5 µM A-485 (Tocris, 6387) was added and the reaction was allowed to equilibrate at room temperature for 10 min. Next, 1 volume of phase-separation buffer containing 5 µM A-485 (Tocris, 6387) was added and the mixture was incubated for 10 min at room temperature before transferring the suspension to a well of a passivated 384-well microscopy plate.

Microtubule in vitro polymerization
To generate stabilized microtubule seeds for nucleation of microtubule networks in vitro, 22 µl of 5 mg ml −1 tubulin protein (Cytoskeleton, T240) was mixed with 2 µl of 5 mg ml −1 TRITC-labelled tubulin protein (Cytoskeleton, TL590M) and 1 µl of 5 mg ml −1 Biotin-XX-labelled tubulin (Pursolutions, 033305). All tubulin storage solutions were prepared in BRB80 (80 mM K-PIPES, pH 6.9 (Sigma-Aldrich, P6757), 1 mM MgCl 2 (Sigma-Aldrich), 1 mM EGTA (Merck, 324626)). The tubulin mixture was centrifuged at 4 °C for 15 min in a tabletop centrifuge (Eppendorf, 5424R) at 21,000g. To 22.5 µl of the resulting supernatant, 2.5 µl 10 mM guanylyl-(alpha,beta)-methylene-diphosphonate ( Jena Biosciences, NU-405) was added to a final concentration of 1 mM and the resulting solution was incubated at 37 °C in a water bath in the dark for 30 min. The resulting seeds were sheared using a 22-gauge needle Hamilton syringe (Sigma-Aldrich, 20788). The resulting suspension was stored in the dark. Seeds were prepared freshly each day and used for subsequent experiments.
To visualize chromatin droplets along with polymerized microtubules, imaging chambers were constructed as described previously 62 . Before Pluronic F-127 passivation, each silanized 24 × 60 mm cover glass was cut into a 24 × 25 mm (top piece) and 24 × 35 mm (bottom piece) using a diamond-tipped steel scribe (Miller, DS-60-C). The bottom coverslip was mounted into the support slide and fixed in place with preheated VALAP (1 part Vaseline (Sigma-Aldrich, 16415), 1 part lanolin (Sigma-Aldrich, L7387), 1 part paraffin wax (Sigma-Aldrich, 327204); all parts by weight). In the central region of the bottom cover glass, a two-well silicon culture-insert (Ibidi, 80209) was attached and 40 µl of nucleosome array droplet suspension was added to one of the wells. The suspension was sedimented for 15 min in a humidified chamber at room temperature. Next, 32 µl of buffer was removed from the well, the culture-insert was removed from the coverslip and 20 µl soluble tubulin mix was added to the remaining chromatin droplet suspension. The top coverslip was added on top with the passivated side facing the reaction mixture, and the droplet was allowed to spread fully (~10 s). The sample was next sealed with VALAP to prevent evaporation and thermal streams within the liquid film. This procedure yielded a liquid film of ~30 µm thickness. The assembled imaging cell was transferred to the incubated stage of a customized Zeiss LSM980 microscope combined with the Airyscan2 detector, using a ×63 1.4 NA oil-immersion DIC plan-apochromat objective (Zeiss). The sample was incubated at 37 °C for >30 min before images were recorded.

CLEM and electron tomography
HeLa cells stably expressing H2B-mCherry were cultured on carboncoated Sapphire discs (0.05 mm thick, 3 mm diameter; Wohlwend). To enrich prometaphase cells, cells were synchronized to G2-M using a double thymidine (Sigma-Aldrich) block for 16 h with 2 mM thymidine each and a subsequent RO3306 (Sigma-Aldrich) block for 6 h with 8 µM RO3306. During the RO-3306 block, the cells were treated with 5 µM TSA for 3 h before RO3306 washout. RO3306 was washed out by rinsing the cells three times with prewarmed imaging medium. The cells were observed on a customized Zeiss LSM780 microscope, using a ×20 0.5 NA EC PlnN DICII air objective (Zeiss). Then, 25 min after the RO3306 washout, most of the cells reached prometaphase (based on DIC and chromatin morphology), and the cells were subsequently processed for electron microscopy analysis. Immediately before freezing, the cells were immersed in cryoprotectant solution (imaging medium containing 20% Ficoll PM400; Sigma-Aldrich), and instantly frozen using a high-pressure freezing machine (HPF Compact 01, Wohlwend).
Frozen cells were freeze-substituted into Lowicryl HM20 resin (Polysciences, 15924-1) using freeze-substitution device (Leica EM AFS2, Leica Microsystems) as follows: cells were incubated with 0.1% uranyl acetate (UA) (Serva Electrophoresis, 77870) in acetone at −90 °C for 24 h. The temperature was increased to −45 °C at a rate of 5 °C h −1 and the cells were incubated for 5 h at −45 °C. Cells were washed three times in acetone at −45 °C and then incubated in increasing concentrations of resin Lowicryl, HM20 in acetone (10, 25, 50 and 75% for 2 h each) while the temperature was further increased to −25 °C at a rate of 2.5 °C h −1 . Cells were then incubated in 100% resin Lowicryl, HM20 at −25 °C for a total of 16 h, changing the pure resin solution after 12 h, 14 h and 16 h. The resin was polymerized under ultraviolet light at −25 °C for 48 h. The temperature was increased to 20 °C (5 °C h −1 ) and ultraviolet polymerization was continued for another 48 h.
After resin polymerization, sections of 250 nm thickness were cut with an ultramicrotome (Ultracut UCT; Leica) and collected on copperpalladium slot grids (Science Services) coated with 1% formvar (Plano). The sections were first observed on the Zeiss LSM710 microscope, using a ×40 1.4 plan-apochromat oil-immersion objective (Zeiss). H2B fluorescence was used to choose cells and subcellular ROIs for tomography. The sections were then post-stained with 2% UA in 70% methanol at room temperature for 7 min and lead citrate at room temperature for 5 min. The sections were observed by Tecnai F20 transmission EM (200 kV; FEI). For tomography, a series of tilt images was acquired over a −60° to +60° tilt range with an angular increment of 1° using Serial EM software 63 at a final xy pixel size of 1.14 nm. Tomograms were reconstructed using the R-weighted back projection method implemented in the IMOD software package 64 .
Images of fields of cells to determine the apoptotic index were recorded on an inverted Axio Observer Z1 with a sCMOS camera equipped with a ×20/0.5 plan-neofluar air objective, controlled by ZEN Blue. For detection of Hoechst 33342 fluorescence, a Ex 377/350 nm, Em 447/60 nm filter was used; for detection of pSIVA fluorescence, a Ex 470/40 nm, Em 525/50 nm filter was used; and for detection of PI fluorescence, a Ex 530/75 nm, Em 560/40 nm was used.
For FRAP experiments, selected image regions were bleached using a laser intensity 200-fold higher than the laser intensity used for image acquisition, and each bleached pixel was illuminated 20 times with the pixel dwell time used for acquisition (1.79 µs). Images were acquired every 25 ms for the course of the experiment.
Images of nucleosome array droplets and in vitro polymerized microtubules were recorded on a customized Zeiss LSM980 microscope combined with the Airyscan2 detector, using the ×40 or ×63 1.4 NA oil-immersion DIC plan-apochromat objectives (Zeiss), operated by ZEN3.3 Blue 2020 software. For mounting 384-well microscopy plates, the Pecon Universal Mounting Frame KM adapter was used. For mounting microtubule imaging slides, a custom aluminium mounting block for 24 mm × 24-60 mm coverslips was used (IMP/IMBA workshop and BioOptics).

Microinjection experiments
Live-cell microinjection experiments were performed using a FemtoJet 4i microinjector in conjunction with an InjectMan 4 micromanipulation device (Eppendorf). All microinjections were performed using pre-pulled Femtotips injection capillaries (Eppendorf). The microinjection device was directly mounted onto a customized confocal Zeiss LSM780 (Fig. 2c and Extended Data Fig. 8a-h) or a customized Zeiss LSM980 microscope (Fig. 2a,e and Extended Data Figs. 3a, 4a, 7a,c,e,g,  8k,l,o, 9a,c,l,k and 10a,c) with live-cell incubation (37 °C, 5% CO 2 ). For all microinjections, cells were cultured in µ-Dish 35 mm high-wall imaging dishes with a polymer or glass bottom (Ibidi) to reach near-confluency on the day of the injection.
For injection of nucleosome arrays (Extended Data Fig. 8k), 1.5 µl injection buffer (50 mM K-HEPES, pH 7.4, 25% glycerol buffer) was added to 1.5 µl unmodified and 1.5 µl acetylated nucleosome array solution (1.3 µM final concentration for each nucleosome in the injection solution). Microinjection of mitotic cells was performed using injection settings of 180-190 hPa injection pressure, 0.35 s injection time and 85 hPa compensation pressure.

Image analysis DNA congression analysis.
To quantify DNA congression to the spindle equator in live cells, the DNA distribution along a line profile parallel to the pole-to-pole axis was measured (for Fig. 1a,b, 7.06 µm width, 22.5 µm length; for Extended Data Fig. 5c,d, 12.04 µm width, 22.5 µm length). Along each line profile, the accumulated DNA density was measured in the central 5 µm interval around the centre position between the poles (determined by highest SiR-tubulin peak intensities) and divided by the total DNA density along the entire profile, after subtraction of the extracellular background fluorescence. Each line profile represents an average intensity projection of z-slices around the pole-to-pole axis (for Fig. 1a,b, 2.4 µm range; for Extended Data Fig. 5c,d, 0.75 µm range).

Chromatin/DNA compaction analysis.
To quantify DNA density, in a central z-section of a mitotic cell (determined by visual inspection on the basis of the highest SiR-tubulin staining intensity at the poles (Fig. 1a,c and Extended Data Figs. 4a-d and 5c,e) the DNA channel was denoised using a Gaussian blur filter (σ = 2) and thresholded using the Otsu dark method in Fiji. The resulting binary mask was converted into a selection and the DNA mean fluorescence within this ROI was measured. All data points were normalized to the mean of unperturbed control cells.
In STLC-treated cells used for AluI microinjection experiment (Fig. 2a,b,e,f and Extended Data Figs. 7e-h and 8g,h), the DNA density was measured in line profiles. In a single z-slice, a line profile (3 pixels wide, 2 µm long) through a chromosome parallel to the imaging plane (before AluI injection), a chromatin droplet (20 min after AluI injection) or the diffuse chromatin mass in TSA treated cells (20 min after AluI injection) was measured. The mean histone 2B fluorescence intensity in a 200 nm interval around the peak value was measured. Values were normalized to the mean of the non-injected control (Fig. 2b,f and Extended Data Figs. 7f and 8h) or the non-injected condensin-degraded control (Extended Data Fig. 7h).
In cells subjected to G2-mitosis induction (Extended Data Fig. 8a-h), the YFP channel was denoised using a Gaussian blur filter (σ = 2) and thresholded using the Otsu dark method in Fiji. The resulting mask was converted into a ROI and the histone 2B-YFP mean fluorescence recorded within this ROI. The values were normalized to the mean of the G2 measurements.
DNA displacement analysis after mitotic entry with monopolar spindle configuration. To quantify DNA displacement away from the spindle pole after mitotic entry in the presence of STLC to induce a monopolar spindle geometry, the DNA/chromatin distribution was measured along a radial line profile having the centre of mass of SiRtubulin fluorescence as a seeding point. From videos of mitotic entry, the time point corresponding to 20 min after prophase entry was used for measurement (Extended Data Fig. 2a-c). The measurement was performed in a maximum intensity z-projection of 2 confocal slices around the centre of mass of SiR-tubulin fluorescence, indicating the centre of the monopolar spindle (z-offset of 2.5 µm). A radial line profile was drawn, using the ImageJ Plugin Radial Profile Angle (v.1-14-2014 by P. Carl and K. Miura, based on Radial Profile from P. Baggethun), with an integration angle Θ = 180°. Along the integrated histone 2B fluorescence around the monopole, the fluorescence signal within the inner 30% distance of the pole to the outer edge of chromatin distance (~99% of cumulative histone 2B fluorescence) was divided by the outer 70%. The resulting value represents the relative amount of chromatin detectable in pole-proximal regions.

Kinetochore displacement analysis.
In z-projections of Airyscan or confocal slices (7 Airyscan z-sections with a z-offset of 0.15 µm for Extended Data Fig. 1f, 2 confocal z-sections with a 2.5 µm z-offset for Extended Data Fig. 2a-c), the histone 2B channel was denoised using a Gaussian blur filter (σ = 2) and thresholded using the Otsu dark method in Fiji and converted into a mask. Kinetochores were detected after denoising the CENP-A channel using a Gaussian blur filter (σ = 1) using the Find Maxima function in Fiji (strict, excluding edge maxima, prominence = 100 (Extended Data Fig. 1g), prominence = 2,000 (Extended Data Fig. 2e)). Of the total number of detected kinetochores, the fraction of kinetochores outside chromatin (>0.5 µm distance to the nearest chromatin mask surface) was quantified manually.
Electron tomography analysis. Analysis of electron tomograms were performed in IMOD/3dmod, v.4.11. Annotation of all structures was performed in the Slicer Window view, using running z average intensity projections to increase contrast (microtubule annotation: projection of 10-20 z-slices, chromatin annotation: 35-50 z-slices). To account for the loss in image sharpness towards the top and bottom of the recorded tomograms, for all of the tomograms, only the centre section of the recorded volume was analysed (tomograms with 120 to 150 slices, no annotation in the top and bottom 20 slices). For microtubule annotation, a zoom factor of 0.8 to 1.0 was used. For chromatin annotation, a zoom factor of 0.35-0.45 was used. Assignment of microtubules was based on ultrastructural morphology, and assignment of chromatin boundaries was performed on the basis of local grain size, considering the transition of a large-grained, coarsely interspersed particle-containing area (cytoplasm with ribosomes; typical coefficient of variation (CV) > 0.4) to a fine-grained, finely interspersed particle containing area (chromatin with nucleosomes; typical CV < 0.3 for control, CV < 0.4 for TSA) as the chromosome surface. The electron density as a determining factor of a chromatin surface was only used as a secondary direction as TSA treatment perturbs the chromatin density/morphology and therefore decreases the annotation accuracy. On average, an annotation landmark was set every 5-10 nm for both microtubules and the chromatin surface. After manual annotation of microtubules and a chromatin surface, a model of a meshed chromatin surface was generated by averaging five consecutive annotated slices to increase surface smoothness. Microtubule segment length was measured within the cytoplasm and chromatin-internal regions and normalized to the total volume of the respective domain.

Mitotic duration measurements in the absence or presence of TSA.
Fields of asynchronous cells stably expressing histone 2B-mRFP were imaged for up to 4 h. The timing of nuclear envelope breakdown and anaphase onset was determined by visual inspection of chromatin morphology.

FRAP analysis of chromatin mobility.
To measure FRAP, raw data measurement, background subtraction, data correction and normalization were performed according to ref. 65 . The measurement of FRAP ROI and background fluorescence signal was performed directly in ZEN software. Total chromatin fluorescence per time-lapse movie was measured using the Time Series Analyzer v.3 ( J. Balaji).

FRET analysis of aurora B-FRET biosensor.
To determine the mitotic state of control and AluI-injected cells at the G2 to M transition (Extended data Fig. 8a-h), CFP, FRET and YFP signals of the histone 2B fused Aurora B-FRET biosensor 66 were recorded at the same time. To quantify FRET efficiency, a custom Fiji script was used. In average-intensity projections of 9 z-slices (1 µm each), a nuclear mask was generated using the YFP fluorescence, after denoising using a Gaussian blur filter (σ = 5), by thresholding using the Otsu dark method. For each time point, the background for the FRET and CFP channel was measured in an extracellular ROI (square of around 1 µm × 1 µm). After background subtraction, FRET and CFP signals within the nuclear ROI were denoised using a Gaussian blur filter (σ = 5) and a FRET/CFP ratio was calculated. The resulting ratiometric image was depicted using the Fire lookup table in Fiji, at an interval of 0 to 1.4 for FRET/CFP.

Measurement of Ki-67 surface confinement.
To determine the surface confinement of Ki-67 to the surface of mitotic chromatin under various conditions, the distribution of Ki-67 was measured along line profiles across the chromatin-cytoplasm boundary.
In mitotic cells arrested in taxol, chromosomes were identified, which were perpendicular to the imaging section in 3 consecutive slices (z-offset between Airyscan slices of 0.15 µm). The centre slice was used for analysis. A line profile of 10 px width was manually drawn through one chromatid in a single Airyscan section, orthogonal to the surface of the chromatid and near-orthogonal to the longer axis of the chromosome. The Ki-67 fluorescence was measured along the line and aligned to the first peak in fluorescence intensity (Fig. 2e,f). For measuring Ki-67 distribution across the chromatin-cytoplasm boundary in cells after AluI-digestion of mitotic chromatin, the line profile was drawn perpendicular to the chromatin droplet surface, otherwise following the same measurement parameters as described above. To quantify the surface confinement of Ki-67, the fluorescence value along the line profile for the highest value of the control (surface) was divided by the value in the centre of the chromatid for control (inside). The same positions along the line profiles were used to calculate the ratios for TSA-treated and AluI-digested cells. (Fig. 2f,g).
To measure the Ki-67 distribution on mitotic chromatin after degradation of SMC4 or degradation of SMC4 and TSA treatment, a line profile of 10 px width was manually drawn across the chromatin surface in a central (based on visual inspection) Airyscan z-section (z-offset, 0.15 µm). The Ki-67-fluorescence was measured along these profiles and aligned around the 50% maximum value of the histone 2B fluorescence, indicative of the edge of the chromatin mass (Extended Data Fig. 8e,f).
Macromolecule partitioning relative to chromatin. In images of microinjected nucleosome arrays into metaphase cells (Extended Data Fig. 8k), the DNA channel was denoised using a Gaussian blur filter (σ = 2) and thresholded using the Otsu dark method in Fiji. The chromatin mask was converted into a ROI by using the Analyze Particles function in Fiji. For the two injected nucleosome arrays of different colours, the mean fluorescence intensity was measured in the DNA ROI and divided by the fluorescence intensity in the cytoplasm measured in a circular ROI of ~5 µm diameter, with at least 1 µm distance to the chromosomes.
The TRITC-tubulin partition coefficient (Fig. 3a,b) was measured >10 min after microinjection in nocodazole-arrested cells. At the surface of a chromosome at the edge of the chromosome cluster, mean fluorescence intensity of TRITC was measured along a line profile (5 px width, 1.3 µm length) orthogonal to the chromosome surface. Values 1 µm apart within and outside of chromatin were divided, resulting in the measured equilibrium partition coefficient.
The distribution of differently expressed DsRed fusion proteins (Fig. 3c,d) was measured using a custom Fiji script. A chromatin mask was generated using thresholding (default method, Fiji), after Gaussian blur filtering (σ = 2) the chromatin channel. The mask was then shrunk by 0.2 µm and the mean fluorescence was measured within the mask. Cytoplasmic fluorescence was measured within a ring (width, 0.5 µm), generated by extending the mask by 1.5 µm. Background mean fluorescence was measured within a small circular region outside the cell and subtracted from chromatin and cytoplasmic values. Partitioning coefficients were obtained by dividing chromatin by cytoplasmic fluorescence values. Measurements were performed on the central z-slice.
The distribution of microinjected differently charged GFPs (Extended Data Fig. 9a,b) and differently modified FITC-labelled dextran fractions (Extended Data Fig. 9c,d) was measured analogously along line profiles orthogonal to the metaphase plate at the orthogonal surface of the chromatin mass (5 px width, 2.5 µm length). Values 1 µm apart within and outside of chromatin were divided, resulting in the measured equilibrium partition coefficient.
The distribution of soluble TRITC-tubulin dimers (Fig. 3e,f), differently charged GFPs (Extended Data Fig. 9m,n) and differently modified FITC-labelled dextran fractions (Extended Data Fig. 9o,p) within nucleosome array droplets in vitro was measured by determining the mean fluorescence intensity in the droplet in a central section of the droplet, in a small circular region covering ~25% of the droplet area in the respective section. In a large, rectangular buffer ROI without droplets, the mean fluorescence in the buffer was measured and the partition coefficient was calculated for each field of droplets.

Chromatin droplet distribution analysis in cells.
Chromatin distribution in pole-peripheral regions was measured at 36 min after nocodazole washout. The spindle-pole position was determined as the z-slice with highest tubulin signal (SiR-tubulin (Fig. 4a) or eGFP-α-tubulin (Extended Data Fig. 10a,c)) and the total histone 2B ( Fig. 4a and Extended Data Fig. 10a,c) and eGFP-CENPA fluorescence intensity in circular ROIs (r = 5 µm) in maximum intensity projections of 5 z-slices (1 µm offset between slices) around the central section was subtracted from the total cellular fluorescence in maximum intensity projections (the cell outline was determined by DIC channel). Each dataset was normalized to the mean of the total fluorescence intensities of all cells.

Microtubule density distribution in vitro.
After deconvolution in ZEN, Airyscan images of fields of microtubules were denoised using a Gaussian blur filter (σ = 2). The background was then removed using the subtract background function in Fiji, using a rolling ball radius of 740 px. The image was converted into 8-bit, and thresholded using the Auto Local Threshold (v.1.10.1) plugin in Fiji, with the Phansalkar method and a radius of 100. The resulting binary image was skeletonized using the Skeletonize function in Fiji. The resulting skeleton length was measured. The chromatin channel was denoised using a Gaussian blur filter (σ = 2) and thresholded using the Otsu method in Fiji. The resulting binary image was transformed into a ROI using the Analyze particles function (size range, 5-infinity). Microtubule skeleton lengths were measured within chromatin ROIs and the surrounding buffer, and the ratio of total segment length in chromatin/buffer was calculated per image.

Validation of auxin-inducible degradation of SMC4-mAID-HaloTAG in live cells.
In images of fields of live cells stained with Oregon-Green488 HaloTAG ligand, the total fluorescence intensity in average intensity projection was recorded in Fiji. To normalize the values to the cell count, the ratios of HaloTAG fluorescence over SiR-DNA intensity were calculated, and the values were normalized to the mean of the Hela WT cells.

Acetylated histone fluorescence intensity measurements.
After indirect immunofluorescence analysis of the respective histone-acetyl marks, a mask of the DNA channel was generated by thresholding using the Otsu method in Fiji. Next, the obtained binary image was converted to ROIs using the Analyze particles function (size range, 5-infinity). Mitotic and interphase cells were differentiated manually on the basis of chromatin morphology. Within the obtained ROI per cell, the mean fluorescence intensity of the acetyl mark was recorded, the extracellular background measured in a rectangular ROI was subtracted and a ratio of the histone mark over DNA mean intensity was generated to account for changes in DNA compaction between the different conditions. For comparison of untreated interphase and mitotic cells, obtained ratios were normalized to the mean of untreated interphase cells (Extended Data Fig. 3a,b). For comparison of untreated and TSA-treated mitotic cells, ratios were normalized to the mean of untreated mitotic cells (Extended Data Fig. 3c,d). For comparison of p300-expressing mitotic cells, ratios were normalized to the mean of the mock-plasmid transfected mitotic cells (Extended Data Fig. 5a,b).

Measurement of cyclin B1 stain.
In interphase mitotic cells, determined by the absence or presence of cell rounding, in the central section of the recorded z-stack (7 z-slices, 1 µm each, central section manually determined), the cell outline in DIC images was used to determine a ROI per cell. The total fluorescence intensity of the cyclin B1 fluorescence within this ROI was recorded in Fiji. The values were normalized to the mean of the cyclin B1 stain in interphase cells.

Quantification of the apoptotic index.
In fields of cells, the DNA channel was denoised using a Gaussian blur filter (σ = 2) and thresholded using the Otsu dark method in Fiji, and one ROI per cell was derived using the Analyze Particle function in Fiji. Within each ROI, the mean fluorescence signal of pSIVA and PI was quantified and any cell with a mean fluorescence signal of greater than 1.2× the median of all untreated control cells was considered to be positive for the respective marker and therefore dead/apoptotic. The fraction of all cells scoring positive for pSIVA and PI stain reflects the apoptotic index (Extended Data Fig. 3e,f).
Quantification of DNA damage foci by γH2A.X staining. A confocal stack of 21 slices with a z-offset of 0.5 µm was converted into a maximum intensity projection. The z-projection was denoised using a Gaussian blur filter (σ = 2), and γH2A.X foci were detected using the Find Maxima tool in Fiji (strict detection, 3,000 prominence) within a segmented chromatin mask obtained by thresholding the DNA channel using the Otsu dark method in Fiji (Extended Data Fig. 3g,h).

Determination of coefficient of variation of chromatin droplets in vitro.
In entire fields of chromatin droplets or nucleosome array solution (Extended Data Fig. 8c) (both recorded ~3 µm above the cover glass using laser powers adjusted to the unmodified condition), the mean fluorescence intensity and s.d. was measured, and the CV (CV = α µ −1 ) was calculated.
Airyscan processing and deconvolution. Raw images recorded using the Airyscan2 detector using AS-SR or multiplex airyscan modes were processed using ZEN3.3 Blue 2020 software.
Fiji version details. The Fiji-integrated distribution/version used for analyses in this study was ImageJ v.1.53c, using Java v.1.8.0_66 (64 bit), with in part custom ImageJ plugins as indicated.

Statistical analyses and data reporting
No statistical methods were used to predetermine sample size. To test significance, the robust, nonparametric Mann-Whitney U-test was used. The statistical test used are indicated and the exact P values are been provided wherever possible. In some cases, the P values calculated were below the precision limit of the data type and algorithm used. In those cases, this has been noted in the legend and an upper bound was given. NS, P > 0.05; *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, ****P ≤ 0.0001. Figure 1a

Reporting summary
Further information on research design is available in the Nature Research Reporting Summary linked to this paper.

Data availability
Raw microscopy data are available and will be provided by the corresponding authors on request, given the large file sizes that are involved. No restrictions apply to the availability of the microscopy data. Source data are provided with this paper.

Article
Extended Data Fig. 3 | Histone acetylation during cell cycle progression and effect of TSA. a,b, Immunofluorescence analysis of histone acetylation in interphase and mitosis. a, HeLa cells were fixed and stained with antibodies against different acetylated histones as indicated. b, Quantification of histone acetylation by immunofluorescence as in a, by the ratio of antibody fluorescence to DNA reference staining by Hoechst 33342. For each acetylated histone, all data points were normalized to the mean of interphase cells. n = 20 for H2B-Ac, Interphase, n = 20 for H3-Ac, Interphase, n = 20 for H4K16-Ac, Interphase, n = 20 for H2B-Ac, Mitosis, n = 20 for H3-Ac, Mitosis, n = 20 for H4K6-Ac, Mitosis. Bars indicate mean; significance was tested by a two-tailed Mann-Whitney test (H2B-Ac, P = 2.117x10 −7 ; H3-Ac, Mitosis, P = 4.72x10 −4 ; H4K16-Ac, P = 1.451x10 −11 ). c,d, Histone acetylation in mitotic cells after TSA treatment. c, HeLa cells were treated with TSA for 3 h, fixed, and histone acetylation analysed by immunofluorescence as in a. d, Quantification of histone acetylation as in b for mitotic cells 3 h after TSA treatment. For each antibody, all data points were normalized to the mean of control metaphase cells. n = 20 for H2B-Ac, control, n = 20 for H3-Ac, control, n = 20 for H4K16-Ac, control, n = 20 for H2B-Ac, TSA, n = 20 for H3-Ac, TSA, n = 20 for H4K16-Ac, TSA. Bars indicate mean; significance was tested by a two-tailed Mann-Whitney test (H2B-Ac, P = 1.451x10-11; H3-Ac, P = 1.451x10 −11 ; H4K16-Ac, TSA, P = 1.451x10 −11 ). e,f, Analysis of apoptosis after TSA treatment. e, Fields of cells stained with Hoechst 33342, pSIVA, and PI to detect apoptotic cells were untreated (control) or treated for 3 h with 500 nM or 5 µM TSA, or 5 µM anisomycin as positive control. f, Quantification of apoptotic index after drug treatment of asynchronous cells, each dot representing a field of cells shown in e, with n = 4574 for untreated, n = 4926 for 500 nM TSA, n = 4653 for 5 µM TSA, and n = 4188 anisomycin cells in total. Bars indicate mean, significance was tested by a two-tailed Mann-Whitney test (500 nM TSA, P = 2.971x10-4; 5 µM TSA, P = 0.078; anisomycin, P = 1.923x10 −7 ). g,h, Immunofluorescence analysis of yH2A.X DNA damage foci in mitosis after treatment with 5 µM TSA or the DNAdamaging agent neocarzinostatin (NCS) as positive control. g, Mitotic cells were stained for yH2A.  -p300. a, b, Immunofluorescence analysis of histone acetylation in mitotic cells overexpressing p300 histone acetyltransferase or catalytically dead p300(D1399Y). a, Cells were transfected with a plasmid coding for eGFP-p300 or eGFP-p300(D1399Y) as indicated and fixed after 48 h. Histone 2B acetylation was analysed by immunofluorescence. DNA was stained with Hoechst 33342. b, Quantification of histone acetylation in metaphase cells as in a. Data points were normalized to the mean of mock-transfected mitotic cells. n = 20 for mock-transfected, n=26 for p300, n = 20 for p300(D1399Y). Bars indicate mean; significance was tested by a two-tailed Mann-Whitney test (P = 3.57x10 −13 ). c-e, Analysis of chromatin density and chromosome congression to the spindle centre in cells after SMC4-AID-Halo degradation (ΔCondensin) and overexpression of p300 or catalytically dead p300(D1399Y). c, Cells were transfected with a plasmid coding for eGFP-p300 or eGFP-p300(D1399Y) as indicated and analysed by live-cell imaging after 48 h. DNA was stained with Hoechst 33342 and microtubules stained by SiR-Tubulin to identify mitotic cells with bipolar spindles. Projection of 5 Z-sections. d, Quantification of chromosome congression by the fraction of chromatin localizing to the central spindle region. n = 20 for mock-transfected, n = 24 for ΔCondensin+p300, n=20 for ΔCondensin+p300(D1399Y). Bars indicate mean; significance was tested by a two-tailed Mann-Whitney test (P = 1.451x10 −11 ). e, Quantification of chromatin density in cells treated as in c. n = 20 for mock-transfected, n = 24 for ΔCondensin+p300, n = 20 for ΔCondensin+p300(D1399Y). Bars indicate mean; significance was tested by a two-tailed Mann-Whitney test (P = 7.86x10 −13 ). Biological replicates: n = 2 (a,b); n = 3 (c-e). Scale bars, 5 µm.