NuMA promotes constitutive heterochromatin compaction by stabilizing linker histone H1 on chromatin

28 Heterochromatin has been widely recognized to exert pivotal functions of silencing specific genes 29 and maintenance of genome stability. However, the mechanisms underlying heterochromatin formation 30 and maintenance remain to be fully elucidated. Here, we discovered that the critical mitotic regulator 31 NuMA, as a nucleoskeleton protein, is required for constitutive heterochromatin organization at the 32 level of nucleosomes in the interphase. Depletion of NuMA results in shortened nucleosome repeat 33 length (NRL), dispersed nucleosome clutches and increased chromatin accessibility in heterochromatin 34 regions. Afterwards, epigenetic maintenance and transcription repression in constitutive 35 heterochromatin are disrupted upon NuMA-depletion, particularly the up-regulated transcription level 36 of the non-coding long terminal repeat (LTR) elements, indicating the crucial roles of NuMA in cell 37 differentiation and senescence. We revealed that such functions of NuMA rely on its interaction with 38 linker histone H1, which stabilizes H1’s binding to chromatin and facilitates nucleosome stacking. We 39 provided direct structural evidence of NuMA ’s stabilization effect at the highest spatial resolution of 40 was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted July 11, 2025. ; https://doi.org/10.1101/2025.07.08.663620doi: bioRxiv preprint nucleosomes through in situ cryo-ET. Notably, we found that NuMA oligomerizes into quasi-meshwork 41 in nucleoplasm and highly co-localizes with H1 on the chromatin, providing the organization basis for 42 NuMA as a nucleoskeleton protein in chromatin architecture regulation. Collectively, our findings 43 illuminate the concerted effect of nucleoskeleton protein and linker histone on chromatin compaction at 44 the level of nucleosomes, which unveil a new layer of mechanisms by which nucleoskeleton regulates 45 heterochromatin formation and maintenance. 46

47 Heterochromatin refers to transcriptionally repressive chromatin domains essential for genome 48 integrity and cellular identity. Classified by developmental dynamics and genomic distribution, 49 facultative heterochromatin exhibits cell type or differentiation state-specific silencing patterns, 50 whereas constitutive heterochromatin usually remains permanently condensed throughout the cell cycle 51 and is enriched in pericentromeric regions, telomeres, and non-coding repetitive DNA elements. The 52 assembly and maintenance of heterochromatin relies on sophisticated interplays among histone 53 modifications (H3K27me3 for facultative heterochromatin and H3K9me2/H3K9me3 for constitutive 54 chromatin) and various chromatin effectors, such as chromatin-remodeling complexes and 55 heterochromatin protein 1 (HP1) family proteins 1-5. Despite current progress, the molecular 56 mechanisms underlying heterochromatin establishment and maintenance remains incompletely 57 resolved, with a substantial number of regulatory factors awaiting discovery. 58 Emerging evidences implicate that nucleoskeleton participates in chromatin organization 6-9 and 59 heterochromatin structural regulation10, 11. Nucleoskeleton, also called nuclear matrix, was first termed 60 in 1970s, and coined to describe the filamentous meshwork remained visible in nuclei subjected to high 61 salt extraction and DNase digestion 12. Notably, NuMA (Nuclear Mitotic Apparatus), a 238 kDa 13 62 nucleoskeleton protein14, 15 displays dual-phase localization, spindle pole enrichment during mitosis 63 and the nucleoplasm in interphase 16 at a considerably high abundance (around 10 6 copies per 64 nucleus)17-19. While its mitotic roles in spindle assembly have been well-established 19, 20, its functions 65 in interphase remain enigmatic. Several clues suggest NuMA ’s chromatin regulatory potential. Firstly, 66 NuMA was proved to bind DNA in vitro via its C-terminal domain 21-23. Secondly, previous studies 67 showed that NuMA contributes to proper chromatin decompaction and nuclear shape by directly 68 associating with DNA at mitotic exit 22, 23 , and down-regulation of NuMA accelerates apoptotic 69 destruction of nuclei in MCF-7 cells 24 and perturbs tissue-specific epigenetic modifications in breast 70 epithelial cultures 25. Besides, over-expressed NuMA forms meshwork-like lattice structure in HeLa 71 nuclei, and purified NuMA self-assembles into ‘multi-armed’ oligomers in vitro13, 26 , suggesting its 72 scaffold functions. Significantly, NuMA was reported to exhibit interaction with histone H1.2 in the 73 yeast two-hybrid system26. Considering linker histone H1’s central roles in binding the linker DNA of 74 nucleosome core particles (NCPs) to stabilize nucleosome structure and condense chromatin27, there is 75 great potential for NuMA to participate in heterochromatin regulation. 76 In this study, we thoroughly explored the functions and unveiled the molecular mechanisms of 77 NuMA in interphase chromatin organization by combining super-resolution imaging, high-throughput 78 sequencing, cryo-ET and a variety of biophysical and biochemical assays, and proposed that NuMA 79 promotes constitutive heterochromatin compaction and transcriptional repression of LTR through 80 stabilizing histone H1 on chromatin. 81 was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted July 11, 2025. ; https://doi.org/10.1101/2025.07.08.663620doi: bioRxiv preprint

82 NuMA regulates global 3D genome organization in the interphase 83 To investigate NuMA ’s functions specifically in the interphase, we constructed an auxin-inducible 84 degradation (AID) system28 in the HCT116 cell line given the lethality of NuMA knock-out in mice as 85 demonstrated in previous research29. Through adding auxin at the end of second round synchronization 86 (See Methods for details), the protein level of NuMA within the interphase time range was rapidly 87 reduced (Figure 1A). Based on the AID system, we explored the effect of NuMA-depletion on the 88 nucleus and genome organization during the interphase. 89 Firstly, immunofluorescence (IF) of the nuclear envelope labeled by lamin A showed different 90 degrees of shrinkage after NuMA-depletion (Figure 1B), suggesting that NuMA plays a role in nuclear 91 mechanical maintenance, corroborating earlier findings23. Next, fluorescent in situ hybridization (FISH) 92 of chromosome 2 and chromosome 18 which respectively represent chromosomes with relatively 93 higher and lower gene content showed that NuMA-depletion led to significant increase of chromosome 94 volume (Figure 1C). This prompted us to explore whether the enlargement of chromosome is 95 accompanied by alterations in chromosome compaction. To answer this question, we performed 96 DNaseI digestion assay to assess the global chromatin accessibility. The results showed that the 97 genome became more sensitive to DNaseI digestion after NuMA-depletion (Figure 1D), indicating a 98 more accessible and less compacted chromatin status. Collectively, these results suggested that NuMA 99 is involved in nuclear organization and promotes chromatin compaction in the interphase. To further 100 verify these findings, we also constructed the AID system in the U2OS cell line (Figure S1A) and 101 observed similar phenotypes upon NuMA-depletion, including shrinkage of nuclear envelope (Figure 102 S1B), enlarged chromosomes volume (Figure S1C), and increased chromatin accessibility (Figure 103 S1D). 104 Based on the observed critical role of NuMA in chromatin spatial organization, we performed 105 high-throughput chromosome conformation capture (Hi-C) assay 30 in control and NuMA-depleted 106 HCT116 cells to measure the hierarchical genome architecture alterations (S1E-G). At the whole 107 genome level, while global chromosomal contact frequency did not exhibit significant changes (Figure 108 S1F and S1G), the inter-chromosomal interactions increased (Figure 1E), in line with the enlarged 109 chromosome volume revealed by FISH (Figure 1C) and enhanced chromatin accessibility examined by 110 the DNaseI digestion assay (Figure 1D). Compartment-level analysis demonstrated prominent 111 decompaction in compartment B, with predominant B-to-A transitions over A-to-B (Fig. 1F and 1G), 112 suggesting that NuMA is important for chromatin compartmentalization and chromatin compaction 113 maintenance in B compartments. The significantly elevated ratio of inter-compartment interactions also 114 suggests a weaker segregation of different chromosomal compartments (Figure 1H and 1I). Taken 115 together, these results indicated that NuMA is required for 3D genome organization at both the 116 chromosome and compartment scales, whose depletion induces global chromatin decompaction, 117 particularly for B compartments. 118 NuMA regulates heterochromatin compaction by maintaining nucleosome 119 stacking 120 To quantitatively assess global chromatin decompaction induced by NuMA-depletion, we performed 121 transposase-accessible chromatin using sequencing (A TAC-seq)31 assay in control and NuMA-depleted 122 was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted July 11, 2025. ; https://doi.org/10.1101/2025.07.08.663620doi: bioRxiv preprint HCT116 cells to map the alterations in chromatin accessibility across the genome. We constructed a 123 chromHMM32 chromatin state map which classified the whole genome into 15 distinct categories for 124 analyzing NuMA ’s impact on chromatin states throughout the genome (Figure 2A). We compared the 125 chromatin accessibility of control and NuMA-depleted cells reflected by A TAC-seq fragment length 126 and density in the 15 categories of chromatin, and calculated the nucleosome-repeat length (NRL) as 127 previously described33. According to the changes in chromatin accessibility and NRL in response to 128 NuMA-depletion, the 15 categories of chromatin exhibited three distinct patterns, which we defined as 129 Type 1, 2 and 3 chromatin (Figure 2B, S2A). Type 1 chromatin demonstrated the most significant 130 increase of A TAC-seq reads frequency and reduction of NRL while Type 2 chromatin exhibits slight 131 increase in chromatin accessibility, and neither of Type 2 or Type 3 chromatin showed obvious changes 132 in NRL after NuMA-depletion (Figure 2B, S2A). Significantly, Type 1 chromatin is predominantly 133 comprised of heterochromatic domains enriched with H3K9me3, H3K27me3, and unannotated 134 post-translational modifications, while Type 2 and 3 chromatin encompasses transcription-active and 135 transcription-regulatory regions. Analysis of Hi-C data according to the categorization of Type 1, 2 and 136 3 chromatin also showed that compartment B was significantly concentrated in Type 1 chromatin 137 (Figure S2B), further verifying that NuMA-depletion mainly affects genome 3D organization in 138 heterochromatin (Figure 1F, G). Furthermore, in regions undergoing compartment A/B switch 139 (including A to B and B to A), Type 1 chromatin also accounted for the highest proportion (Figure S2C). 140 Focusing specifically on compartment A/B switches within Type 1 chromatin, we found that the 141 compartment B to A switch and chromatin de-compaction also accounted for the highest proportion 142 (Figure S2D). Collectively, the integrated analysis of A TAC-seq and Hi-C data indicated that 143 NuMA-depletion leads to a more accessible chromatin state and shortening of NRL in heterochromatin 144 regions, suggesting that NuMA regulates heterochromatin compaction and nucleosome spacing. 145 We also conducted Mnase digestion assay and super-resolution imaging of nucleosome clutches34 to 146 confirm the effect of NuMA-depletion on nucleosome spacing and stacking. The results showed that 147 after NuMA-depletion in HCT116 and U2OS cells, chromatin’s sensitivity to Mnase was increased, 148 which confirmed the shortening of NRL and increase of chromatin accessibility (Figure 2C, S2E). 149 Labeling of H2B by immunofluorescence and imaging of nucleosomes in situ by Stimulated Emission 150 Depletion (STED) imaging revealed that NuMA-depletion had a dispersing effect on nucleosome 151 clutches, evident in reduced average area and decreased nearest neighbor distance (NND) between 152 them (Figure 2D, S2F-H), suggesting NuMA ’s effect on spatial arrangement of nucleosomes. As 153 reported in the previous research, larger and denser clutches containing more nucleosomes formed 154 “closed” heterochromatin, while smaller and less dense clutches formed “open” chromatin 34 which 155 aligns with the observation that NuMA mainly affect the accessibility of Type 1 chromatin (Figure 2B, 156 S2B-D). 157 Collectively, these data demonstrated that NuMA-depletion disrupts genome architecture at the 158 nucleosome level characterized by decreased NRL and reduced density of nucleosome spacing. These 159 structural alterations provide us mechanistic evidence that NuMA governs heterochromatin compaction 160 through maintaining nucleosome stacking (Figure 2E). 161 NuMA stabilizes linker histone H1’s binding to chromatin and facilitates 162 nucleosome stacking 163 As the most abundant chromatin-binding protein in eukaryotes, linker histone H1 binds to the entry 164 was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted July 11, 2025. ; https://doi.org/10.1101/2025.07.08.663620doi: bioRxiv preprint and exit sites of DNA on the nucleosome core particles (NPCs) to stabilize nucleosome structure and 165 compact chromatin27, 35, 36. H1 is one of the direct regulatory factors of NRL, and there is a robust linear 166 relationship between H1 stoichiometry and NRL 27, 36, 37 . In 1998, Gueth-Hallonet and his colleagues 167 reported that NuMA exhibits interaction with histone H1.2 in the yeast two-hybrid system26. Therefore, 168 we hypothesized that the impact of NuMA-depletion on genome architecture, especially at the 169 nucleosome level, is closely related to its interaction with H1. 170 The correlation between NuMA and H1 was substantiated by the co-localization of mScarlet-fused 171 NuMA and eGFP-fused H1.1 revealed by live-cell imaging (Figure 3A). Moreover, we found that 172 among the N-terminus (hereinafter abbreviated as NuMA-N), C-terminus (hereinafter abbreviated as 173 NuMA-C) and 200 nm-long coiled-coil domain of NuMA (Figure S3A), only NuMA-C showed 174 co-localization with H1.1 and six other H1 variants in somatic cells 27 (Figure S3B), suggesting that 175 NuMA interacts with H1 through its C-terminus. This finding was also verified by co-IP of endogenous 176 NuMA with H1 (Figure S3C, S3D), overexpressed NuMA with H1 (Figure 3B, S3E), and 177 overexpressed NuMA-C with H1 (Figure S3F). Besides, we mapped the interaction regions of NuMA 178 binding to H1 through charactering the spatial localization by live-cell imaging and correlation by 179 co-IP between overexpressed NuMA-C and H1. We identified two accurate regions of NuMA-C 180 binding to H1, including region aa 1778-1942 and aa 2004-2115 (Figure 3H, S3G, S3H), which we 181 named as H1BD1 (H1 binding domain 1) and H1BD2 (H1 binding domain 2), respectively. Through in 182 vitro GST-pull down experiments, we also proved that NuMA-C directly binds H1 via H1BD1 and 183 H1BD2 (Figure S3I, S3J). 184 NuMA was reported to bind DNA directly in vitro through its evolutionarily conserved last 58 amino 185 acids in the C-terminus21-23. To profile NuMA ’s in situ DNA binding sites across the whole genome, we 186 performed cleavage under target & tagmentation (CUT&Tag)38 experiments using an antibody targeted 187 to NuMA-C. The results showed that NuMA indeed exhibited specific DNA binding sites on the 188 genome and highly co-localized with the peaks of H1 (Figure 3C and 3D). Interestingly, we observed 189 the highest occupancy of NuMA in Type 1 chromatin (Figure S3K), and the degree of colocalization of 190 NuMA and H1 was also the highest in Type 1 chromatin (Figure 3E). These data are consistent with the 191 observation that NuMA-depletion mostly impacts the architecture of heterochromatin (Figure 2A and 192 2B). 193 Intrigued by the dual binding capacity of NuMA with H1 and DNA, we questioned whether NuMA 194 can affect H1’s binding to chromatin and thereby regulate nucleosome stacking. We performed 195 CUT&Tag experiments of H1 in control and NuMA-depleted HCT116-mAID-NuMA cells, and found 196 that the coverage of H1’s binding sites and counts of peaks were both significantly reduced in Type 1 197 chromatin, in clear contrast to those in Type 2 and Type 3 chromatin (Figure 3F and 3G), which is 198 consistent with NuMA ’s occupancy and NuMA ’s co-localization with H1 on genome (Figure 3E). To 199 validate the ability of NuMA promoting H1’s binding on DNA in vitro, we performed electrophoretic 200 mobility shift assay (EMSA), and found that in the presence of NuMA-C, H1’s DNA-binding capacity 201 was significantly strengthened. In contrast, the truncation of NuMA which deleted its two H1BDs was 202 unable to increase the binding of H1 to DNA (Figure 3I and S3L). Taken together, we concluded that 203 NuMA stabilizes linker histone H1’s binding to chromatin through its interaction with H1. 204 Extensive studies have established linker histone H1's nucleosome-stabilizing role in chromatin 205 condensation through linker DNA binding. Besides of in vitro evidences 39, 40 , H1’s preferential 206 was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted July 11, 2025. ; https://doi.org/10.1101/2025.07.08.663620doi: bioRxiv preprint enrichment in heterochromatin and large nucleosome clutches were also wildly recognized34. Therefore, 207 we proposed that NuMA ’s stabilization effect on H1’s binding to DNA is the mechanism by which 208 NuMA regulates chromatin architecture at the nucleosome level. 209 To confirm this hypothesis, in vitro pull-down assay was applied to evaluate NuMA ’s binding 210 affinity to single nucleosomes. Immunoblotting results revealed that NuMA-C could bind to individual 211 nucleosomes in vitro , and nucleosomes containing histone H1 pulled down significantly more 212 NuMA-C (Figure 4A), suggesting the binding preference of NuMA to H1-associated 213 nucleosomes.Then we performed atomic force microscopy (AFM) to characterize NuMA ’s effect on 214 the conformation of nucleosome array (NA). As shown in Figure 4B, the 12× NA adopted stretched 215 beads-on-string conformation, and the addition of H1 led to a condensed structure. Afterwards, the 216 addition of NuMA-C caused compacted 12× NA to form larger condensates in the presence of H1. 217 NuMA ’s enhancement on H1’s compaction effect on 12× NA was also confirmed by the sucrose 218 density-gradient centrifugation result (Figure 4C). Noteworthily, we found that without the presence of 219 H1, NuMA-C showed no significant effects on either the conformation of 12× NA (Figure 4B) or the 220 sedimentation coefficient (Figure 4C), suggesting that NuMA's regulatory effect on chromatin 221 architecture is dependent on its interaction with H1. 222 To elucidate NuMA ’s facilitation effect on nucleosome stacking under conditions close to native 223 states, we conducted advanced cryo-focused ion beam (FIB) milling combined with cryo-electron 224 tomography (cryo-ET) in U2OS cells overexpressing full-length NuMA (hereinafter abbreviated as 225 NuMA-FL) and a truncated variant of NuMA deleting the C-terminus (hereinafter abbreviated as 226 NuMA-dC, Figure S3A). Tilt-series analysis revealed that nucleosomes were more compactly arranged 227 with a significantly reduced nearest neighbor distance (NND) in cells overexpressing NuMA-FL 228 compared to NuMA-dC (Figure 4D, 4E, S4A). This difference observed between the overexpression of 229 NuMA-FL and NuMA-dC under identical experimental conditions precisely underscored the pivotal 230 role of NuMA-C in nucleosome stacking maintenance. At the same time, subtomogram averaging 231 resolved the conformation of NCPs with nanometer resolution at 9.9 Å from NuMA-dC tomograms, 232 allowing proper fitting of histone helices into the electron density map (Figure S4B, S4D, S4E, 4D). In 233 parallel, a chromatosome conformation was resolved at 17.8 Å resolution from NuMA-FL tomograms 234 using the exactly same workflow as NuMA-dC (Figure S4C, S4D, S4F, 4D). Notably, a distinct density 235 was resolved at the DNA entry/exit site, which could be accurately fitted with the H1 globular domain 236 (Figure 4D, S4F). In contrast, no extra density was detected at the H1 linker region throughout the 237 entire processing workflow from NuMA-dC (Figure S4E), highlighting the stabilization influence of 238 NuMA on H1’s binding to chromatin. Collectively, these results demonstrate that NuMA facilitates 239 nucleosome stacking through stabilizing the interaction between H1 and nucleosome core. 240 NuMA oligomerizes into quasi-meshwork organization through its C-termini in 241 vivo 242 Previous studies reported that purified NuMA self-assembles into ‘multi-armed’ oligomers through 243 its C-terminus in vitro and over-expressed NuMA forms lattice structure in HeLa nuclei13, 26, suggesting 244 its structural and supportive functions. Since we have proved NuMA ’s interaction with H1 and 245 regulatory effect on it, we performed super-resolution imaging to delineate the distribution pattern of 246 endogenic NuMA in the nucleoplasm to seek the structural basis of NuMA for its genome organization 247 functions. 248 was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted July 11, 2025. ; https://doi.org/10.1101/2025.07.08.663620doi: bioRxiv preprint To achieve this, we co-labeled the two ends of NuMA with antibodies specifically targeting to 249 NuMA-N and NuMA-C, respectively (Figure 5A). Due to its considerably high abundance 17-19, both 250 confocal imaging and STED super-resolution imaging were insufficient to resolve individual NuMA 251 molecules (Figure S5A). We therefore performed expansion microscopy as previously described 41, 42, 252 which enabled characterization of Atto647N-labeled NuMA-N and Alexa594-labeled NuMA-C in 253 U2OS cells (Figure 5B). In order to unveil the organization pattern of NuMA molecules, we linked the 254 foci of NuMA-N and NuMA-C within its molecular length, i.e. 250 nm 13, and calculated the 255 stoichiometry and NND between NuMA-N and NuMA-C (Figure 5B). As seen in the cropped view, a 256 NuMA-C focus was encircled by 2 ~ 4 NuMA-N foci, which is similar with the pattern observed in 257 vitro13. The average NuMA-N to NuMA-C ratio was 2.808, and the mean NND from NuMA-C to 258 NuMA-N clusters was 107.22 nm in U2OS cells (Figure 5C). These results suggested that NuMA 259 forms a quasi-meshwork organization via oligomerization of NuMA-C. 260 The 2D characteristics of NuMA distribution were further validated in the three-dimensional space 261 by 3D super-resolution imaging (Figure 5D-F). Notably, the mean NND from NuMA-C to NuMA-N in 262 3D images was 217.72 nm in U2OS cells (Figure 5F), which is close to the length of NuMA molecule 263 and indicates that NuMA adopts an extended conformation in the nucleus. To prevent the influence of 264 different dyes on efficiency of fluorescent labeling and imaging, we swapped fluorescent dyes for 265 NuMA-N and NuMA-C, which yielded similar distribution patterns, stoichiometry and NND values 266 (Figure S5B & S5D). We also performed the experiments and analysis in HCT116 cells and the results 267 were similar with those obtained in U2OS cells (Figure S5C & S5E). Collectively, these imaging 268 analyses supported that NuMA oligomerizes into quasi-meshwork organization through its C-termini in 269 vivo. 270 NuMA contributes to epigenetic maintenance of constitutive heterochromatin and 271 repression of LTR expression 272 Since we have found that NuMA ’s regulatory effect on chromatin architecture at the nucleosome 273 level and facilitation effect on nucleosome stacking by promoting H1’s binding to DNA are mainly 274 showed in regions enriched with heterochromatin markers, our subsequent investigations focused on 275 alterations of heterochromatin markers induced by NuMA-depletion. 276 Because of the high abundance of NuMA in nucleus, we utilized expansion microscopy to 277 characterize the localization of NuMA-C and heterochromatin markers, H3K9me3 and H3K27me3, 278 which marks constitutive heterochromatin and facultative heterochromatin, respectively. H3K4me3, the 279 transcriptionally active euchromatin marker, was also labeled as negative control. Among the three 280 epigenetic markers, we only observed an obvious distribution pattern around H3K9me3 speckles for 281 NuMA-C (Figure 5G). Next, as the expression level of all these markers remained stable (Figure S6B), 282 the distribution of H3K9me3 became more dispersed in the nucleoplasm after NuMA-depletion in 283 HCT116 cells, as rated by the heterogenous index (Described in “Immunofluorescence staining” in 284 Methods). In contrast, the distribution of H3K27me3 remained unperturbed upon NuMA-depletion, 285 which is similar to H3K4me3 (Figure 6A). These results imply that while the effect of 286 NuMA-depletion on chromatin architecture and nucleosome stacking is most significant in 287 heterochromatin, its influence on heterochromatin markers is specifically concentrated on H3K9me3, 288 the constitutive heterochromatin modification. Similar results were obtained in NuMA-depleted U2OS 289 cells (Figures S6A and S6C), reinforcing the specificity of NuMA's role in regulation of constitutive 290 was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted July 11, 2025. ; https://doi.org/10.1101/2025.07.08.663620doi: bioRxiv preprint heterochromatin modification. 291 Considering that NuMA maintains nucleosome stacking through its interaction with H1 (Figure 4) 292 and H1 is widely known to be correlated with heterochromatin formation and maintenance 43-46, we 293 assessed the spatial distribution of H1 alongside NuMA-C and the epigenetic markers. We observed a 294 strong spatial correlation between NuMA-C and H1 (Figure 5H), consistent with their co-localization 295 on the chromatin (Figure 3D and 3E). Moreover, NuMA-depletion was found to greatly reduce the 296 co-localization degree between H1 and H3K9me3, but had no effects on H3K27me3 or H3K4me3 297 (Figure 6B), further substantiating NuMA's specific regulatory effect on constitutive heterochromatin 298 modification. 299 After observing the spatial distribution of heterochromatin markers, we applied CUT&Tag to profile 300 the heterochromatin markers, H3K9me3 and H3K27me3, along the whole genome. In untreated 301 HCT116 cells, H3K9me3 accounted for a higher proportion at binding sites that NuMA-C and H1 302 overlapped than H3K27me3 (Figure 6C), which suggested that NuMA ’s effect on H1 is more 303 concentrated in constitutive heterochromatin than facultative heterochromatin. Then we measured 304 changes of the enrichment of H3K9me3 and H3K27me3 on chromatin after NuMA-depletion, and 305 found that the number of down-regulated H3K9me3 peaks was the highest among all of the altered 306 repressive markers (Figure 6D), illustrating that NuMA-depletion mainly causes down-regulation of 307 constitutive heterochromatin modification. Reciprocally, the overlap of NuMA-C and H1 also located 308 more in regions with down-regulated H3K9me3 than regions with up-regulated and unchanged 309 H3K9me3 (Figure 6E and S6D). When summarizing H3K9me3 peaks according to the binding sites of 310 NuMA-C and H1, we found that the down-regulation of H3K9me3 was most significant at regions 311 where NuMA-C and H1 overlapped (Figure 6F), which suggests that NuMA coordinates with H1 in 312 maintenance of constitutive heterochromatin modification. On the contrary, the absence of apparent 313 pattern in the enrichment of H3K27me3 (Figure S6E) and summary of H3K27me3 peaks in binding 314 sites of NuMA overlapping H1 (Figure S6F) also support this conclusion. 315 Previous research reported that linker histone H1 is enriched in the constitutive heterochromatin that 316 silence repetitive elements in mESCs, and that acute deletion of H1 leads to substantial derepression of 317 repetitive element gene expression 45. Then we were intrigued by NuMA ’s potential effects on 318 transcription, especially in regions that NuMA and H1 overlap. RNA-seq analysis revealed that 319 NuMA-depletion minimally affected gene transcription across the whole genome (Figure S6G) with no 320 specific Gene Ontology (GO) enrichment (Figure S6H), which are consistent with that NuMA mainly 321 regulate constitutive heterochromatin with relative low gene content. However, when we turned to 322 non-coding genomic regions, we found that upon NuMA-depletion, the long terminal repeat (LTR) 323 elements accounted for higher proportion in regions with down-regulated H3K9me3 than regions with 324 unchanged H3K9me3 (Figure 6G). Based on this, the transcription level of LTRs in regions that NuMA 325 and H1 overlapped were also altered, which was significantly increased with up-regulated H3K9me3 326 (Figure 6H, 6I). These results indicated that NuMA collaborates with H1 to contribute to epigenetic 327 maintenance of constitutive heterochromatin and repression of LTR expression, strongly suggesting 328 NuMA ’s crucial role in cell fate regulation during the interphase, such as cell differentiation and 329 senescence. 330

331 Based on the data presented, we proposed a hypothetical model that NuMA serves as nucleoskeleton 332 was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted July 11, 2025. ; https://doi.org/10.1101/2025.07.08.663620doi: bioRxiv preprint protein to regulate heterochromatin maintenance and genome organization in the interphase. NuMA 333 interacts with linker histone H1 and stabilizes H1’s enrichment on chromatin to facilitate nucleosome 334 stacking. Besides of its interplay with H1, NuMA also oligomerizes into quasi-meshwork organization 335 through its C-termini. Put together, both NuMA-H1 and NuMA-NuMA interactions are coordinated in 336 promoting constitutive heterochromatin compaction and repression of non-coding LTRs (Figure 7). 337 Beyond the conventional classification of euchromatin and heterochromatin, the 3D hierarchical 338 organization of chromatin introduces an additional regulatory dimension governing fundamental 339 biological processes, such as DNA replication, transcription, and cell fate decision. Meanwhile, 340 chromatin’s hierarchical architecture exhibits profound spatiotemporal coupling with euchromatin and 341 heterochromatin. Genomic regions of Compartment A are enriched with euchromatin epigenetic 342 markers and transcriptionally active, while Compartment B is mostly transcriptionally repressive and 343 enriched with heterochromatin markers 47. Combining Hi-C and A TAC-seq, we found that 344 NuMA-depletion mainly causes B-to-A compartment transition (Figure 1F-I) and increased genomic 345 accessibility of regions enriched with repressive epigenetic markers (Figure 2A-B, S2A-D), 346 demonstrating that NuMA regulates heterochromatin compaction. These insights also significantly 347 enhance our understanding of the nucleoskeleton’s role in shaping chromatin architecture in 348 mammalian cells. 349 Besides of the compartment-level reorganization, the nucleosome level of chromatin also exhibits 350 significant alterations induced by NuMA-depletion, which is reflected by shortened NRL and dispersed 351 nucleosome clutches (Figure 2, S2). Given that linker histone H1 is one of the direct regulatory factors 352 of NRL27, 37, we confirmed the specific interaction between NuMA and H1, and proved that NuMA 353 stabilizes H1’s enrichment on chromatin to facilitate nucleosome stacking (Figure 3, 4 and S3, 4). This 354 finding correlates nucleoskeleton and chromatin architecture at the level of nucleosomes for the first 355 time. Intriguingly, the phenotypes induced by NuMA-depletion, including B-to-A compartment 356 transition, chromatin decompaction, shortening of NRL, dispersed nucleosome clutches and dispersion 357 of H3K9me3 in heterochromatin regions, are similar with those induced by H1 TKO reported 358 previously33, 34, 44, 45 . This observation is also in line with the enrichment of H1 in heterochromatin 359 regions33. Moreover, in contrast to the changes in nucleosome stacking, chromatin structure at the TAD 360 level, manifested by the distribution of TAD length and insulation scores around TAD boundaries, 361 remained unchanged in control and NuMA-depleted cells (Figure S1H and S1I). Along with the highest 362 degree of colocalization between NuMA and H1 in heterochromatin region (Figure 3F) and the lack of 363 conserved sequence features at NuMA binding sites (data not shown), these results all implies NuMA's 364 reliance on H1 for chromatin binding and 3D genome organization. 365 Moreover, in situ cryo-ET results provided direct structural evidence that NuMA stabilizes the 366 binding of histone H1 to DNA and nucleosomes (Figure 4 and S4). The increased NND among 367 nucleosomes observed in cells overexpressing NuMA-dC compared to those expressing NuMA-FL 368 suggests weaker nucleosome stacking and less compact chromatin organization, which are consistent 369 with the phenotypes induced by NuMA-depletion (Figure 1, 2, S1, and S2). In addition, the absence of 370 H1 density in NCPs map from cells overexpressing NuMA-dC further supports the essential role of 371 NuMA-C in stabilizing H1’s association with nucleosomes. Previous researches reported that H1 372 condensates with DNA and nucleosome fibers to promote heterochromatin formation43, 45, 46, 48. Besides, 373 H1 also functions in heterochromatin formation and maintenance through interaction with other 374 proteins which in turn modify chromatin or take part in DNA-based processes, like recruiting histone 375 was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted July 11, 2025. ; https://doi.org/10.1101/2025.07.08.663620doi: bioRxiv preprint methyltransferase Su(var)3-9 to chromatin and providing binding platform for HP1 in Drosophila 376 melanogaster43. Therefore, the regulatory effects of NuMA on heterochromatin are predominantly 377 attributable to its stabilization of histone H1, thereby unveiling a novel layer of mechanisms by which 378 nucleoskeleton proteins regulate heterochromatin formation and maintenance. 379 The formation and maintenance of heterochromatin are critical for early embryonic development and 380 differentiation processes, during which programmed gene silencing ensures precise spatiotemporal 381 control of developmental cues 1-5. Besides of nucleosome spacing tuning mentioned above, H1 also 382 participates in programmed gene silencing in development. For example, when the protein level of the 383 single somatic H1 subtype was reduced to around 20% of the normal content by RNAi, the 384 H1-depleted D. melanogaster larvae wouldn’t develop to adult flies49. Besides, H1 was also reported to 385 contribute to repression of TEs. In D. melanogaster, loss of H1 caused derepression of more than 50% 386 of TEs and only about 10% of protein-coding genes by physically tethering Su(var)3-9 to facilitate 387 methylation of H3K9 43. This function of H1 to contribute to TE silencing by participating in 388 constitutive heterochromatin formation well matches down-regulation of H3K9me3 peaks induced by 389 NuMA-depletion and the higher co-localization proportion of H3K9me3 with overlapped NuMA-C and 390 H1 on genome than H3K27me3. LTR belongs to the retrotransposons of TE family, and is implicated in 391 several critical biological processes, such as generating mutations in evolution. Typically, most TEs are 392 inactive and tightly regulated to ensure genomic stability 50, 51. In 2022, Gao and colleagues reported 393 that LTRs are gradually silenced by stage-specific H3K9me3 establishment in human early embryos 52. 394 In 2023, Liu and colleagues found that derepression of endogenous retrovirus (ERVs), which accounts 395 for ~97% of LTR, is associated with epigenetic derepression of H3K9me3 and contributes to cellular 396 senescence53. Since the establishment and maintenance of H3K9me3 are critical for repression of LTRs, 397 NuMA ’s compaction effect on constitutive heterochromatin indicates its nonnegligible roles in the 398 maintenance of genomic stability and cell fate decision. Through interrogation of public database 54-56, 399 we found that NuMA’s expression level is stage-specifically elevated during embryogenesis (data not 400 shown), providing orthogonal validation of our proposition that NuMA plays important roles in cell 401 fate decision, rather than merely functioning as a mitotic regulator. 402 Dispersed nucleosome clutches induced by NuMA-depletion also indicates a potential role for 403 NuMA in cellular differentiation and development. The concept of nucleosome clutch was first put 404 forward in 2015 34. Using STORM imaging, the authors found that nucleosomes are arranged in 405 heterogeneous groups of around 20-100 nm diameter along the chromatin fiber. Importantly, a 406 correlation was discovered between the spatial distribution, size, compaction of nucleosome clutches 407 and cell pluripotency. For example, ground-state stem cells have lower-density clutches containing on 408 average only a few nucleosomes than differentiated ones. Therefore, the dispersion and decrease of 409 nucleosome clutches induced by NuMA-depletion (Figure 2D & S2H) and increase of NND among 410 nucleosomes induced by overexpressing of NuMA-dC (Figure 4E) indicate that NuMA has the 411 potential in regulation of differentiation and development processes. 412 In summary, we have provided comprehensive evidence proving that NuMA promotes constitutive 413 heterochromatin compaction and consequently represses LTR expression by stabilizing H1 on 414 chromatin in the interphase. Besides, the NuMA meshwork is probably another mechanism by which 415 it’s involved in chromatin organization, like imposing spatial constraints on chromatin topology. 416 There are some intriguing questions awaiting to be answered. For instance, a few recent studies have 417 was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted July 11, 2025. ; https://doi.org/10.1101/2025.07.08.663620doi: bioRxiv preprint demonstrated that H1 can condensate with DNA in vitro through its highly disordered C-terminal tail57, 418 and H1 promotes chromatin’s liquid-liquid phase separation (LLPS) in vitro 46, 48, 58. Besides, NuMA-C 419 was also reported to regulate mitotic spindle assembly and structural dynamics via phase separation59, 60. 420 Thus, the mechanism of NuMA ’s stabilization effect on H1 in the interphase is also probably related to 421 phase separation. On the other hand, as an essential mitotic spindle regulator, it is unclear how the 422 different functions of NuMA from the mitotic phase to the interphase are transitioned. One possibility 423 is that post-translational modifications of NuMA control the function transition, like phosphorylation. 424 As a hint, NuMA binds both microtubules and H1 through its C-terminus, and the most important 425 phosphorylation modification of NuMA after entering mitotic phase is also exactly located in H1BD219. 426 Besides 3D genome organization, NuMA probably also contributes to nuclear mechanotransduction 427 because of its unique conformation and oligomerization mode, which would be tempting for further 428 exploration. 429

430 Cloning 431 The sgRNA sequence (5’-GGTGGCGTGGAGTGTCA TCT-3’) targeting to NuMA’s N-terminal 432 locus was generated using the online CRISPR design tool ( http://crispor.tefor.net). The sgRNA 433 sequence (5’-GGTGGGGCCACTCACTGGTA-3’) targeting to NuMA’s C-terminal locus was referred 434 to previous report61. The sgRNA fragments were annealed from commercially synthesized single-strand 435 oligos and then inserted into the pX330-U6-Chimeric_BB-CBh-hSpCas9 vector (Addgene, 42230) 436 using Golden Gate cloning (BbsI-HF, NEB, R3539M). 437 The DNA fragment of NuMA and histone H1 isoforms was amplified from a cDNA library produced 438 from HCT116 cells by reverse transcription. Truncations of NuMA was generated through PCR 439 amplification and then cloned into pEGFP-N1, pET-28a (+) and lentivirus vectors through Gibson 440 assembly. 441 Cell culture and transfections 442 The human colon tumor cell line HCT116 was cultured in McCoy’s 5A medium (Gibco, 16600082), 443 while HEK293 cells and human cancer cell lines U2OS and HeLa were cultured in DMEM medium 444 (Gibco, C11995500BT). All types of medium contain 10% fetal bovine serum (Gibco, 10091148) and 1% 445 penicillin-streptomycin (Gibco, 15140122), and all cells were cultured in a humidified incubator at 446 37 °C with 5% CO2. These cells were tested and found to be free of mycoplasma contamination. Before 447 rapid degradation of NuMA, cells were synchronized to the G1/S boundary by two rounds of blocking 448 as described previously 62, 63 . In the first round of synchronization, cells were treated with 2 mM 449 thymidine (Sigma-Aldrich, T1895) for 15 hours followed by releasement to fresh medium 450 supplemented with doxycycline (1 μ g/ml for HCT116 cells and 5 μ g/ml for U2OS cells, Beyotime, 451 ST039) for 10 hours. Then in the second round of synchronization, cells were treated by 2 μ g/mL 452 aphidicolin (Sigma-Aldrich, A4487) for 15 hours. NuMA-depleted cells were treated with 500 μ M 453 auxin/IAA (Sigma-Aldrich, I2886) for 7 hours at the end of the second blocking period. 454 For transient transfection, cells were passaged to approximately 60% confluency one day ahead. 455 Transfection was performed using Neofect transfection reagent (Neofect, TF20121201) following the 456 manufacturer’s protocol. Approximately 0.5 μ g plasmids were transfected into cells in 12-well plates, 1 457 μ g plasmids were transfected into cells in 35 mm Petri dishes or 6-well plates, 2 μ g plasmids were 458 was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted July 11, 2025. ; https://doi.org/10.1101/2025.07.08.663620doi: bioRxiv preprint transfected into cells in 60 mm dishes and 10 μ g plasmids were transfected into cells in 100 mm dishes. 459 After transfection, cells were cultured for 72 hours until the samples were imaged or co-IP 460 experiments. 461 For over-expression of NuMA and its truncations, transfection was accomplished by lentiviral 462 infection. Lentivirus was prepared referring to officially recommended protocols of Addgene 463 (https://www.addgene.org/protocols/lentivirus-production/). HEK293T cells were passaged to 464 approximately 60% confluency one day ahead in 100 mm plates. 10 μ g lentiviral transfer plasmids, 5 465 μ g lentiviral packaging plasmid psPAX2 (Addgene, 12260) and 7.5 μ g envelope expressing plasmid 466 pMD2.G (Addgene, 12259) were co-transfected. The lentiviral particles were harvested by filtering the 467 supernatant through a 0.45 μ m filter 24, 48 and 72 hours post transfection and either used immediately 468 or stored at -80 °C. 469 Knock-in by CRISPR/CAS9 470 Construction of HCT116 and U2OS cells expressing OsTIR1 was performed as reported previously28. 471 Plasmids AA VS1 T2 CRIPR in pX330 (Addgene, 72833) and pMK243 (Tet-OsTIR1-PURO) (Addgene, 472 72835) were co-transfected and positive clones were screened by puromycin. Then pX330-based 473 CRISPR/CAS plasmid which contains sgRNA targeting to NuMA’s N-terminus and homology-directed 474 repair template plasmid containing mAID and Clover were co-transfected into the Tet-OsTIR1 cells. 475 After around two days of culture, cells were sorted via FACS to isolate single cells expressing the 476 Clover protein for imaging. After two weeks, single-colony expansion was verified using western blots. 477 Similar procedures were performed to generate HCT116 cells with N-terminal Avitag and eGFP and 478 C-terminal HA tag in the NuMA gene locus respectively. 479 Antibodies 480 The antibodies used are listed in Supplementary Table 1. 481 Western-blotting 482 For whole-cell samples, cells were harvested by centrifugation at 200 g for 3 min and lysed in RIPA 483 buffer supplemented with 0.1 mg/mL PMSF for 10 minutes on ice. Cell lysate, immunoprecipitation 484 and pull-down input and product were added with SDS-page loading buffer and boiled at 98 °C for 10 485 minutes. Then samples were run on 10% or 15% polyacrylamide gel and wet-transferred to 0.45 μ m 486 PVDF membrane. Next, the membrane was blocked with 5% skim milk dissolved in TBST buffer 487 (Tris-HCl 10 mM, pH 7.5, NaCl 150 mM, Tween-20 0.05%) for 1 hour at room temperature, incubated 488 with primary antibodies at 4 °C overnight. After washed three times with TBST for 5 minutes at RT 489 with shaking, the membrane was incubated with HRP-conjugated secondary antibodies for 1 hour at RT. 490 Finally, the membrane was washed with TBST for five times and covered with ECL substrate. The blot 491 images were acquired with Amersham Imager 600 (GE). 492 Immunofluorescence staining 493 Cells were fixed with 4% paraformaldehyde in 1× PBS for 10 minutes and permeated with 0.5 % 494 Triton-100 in 1× PBS for 30 minutes at RT. Then samples were blocked with blocking buffer (1× PBS, 495 BSA 5%, Triton-100 0.5 %) for 1 hour at RT, and incubated with primary antibodies diluted in blocking 496 buffer at 4 /i2 overnight. The next day, samples were washed five times with 1× PBS for 5 minutes at 497 RT with shaking and incubated with dye-conjugated secondary antibodies diluted in blocking buffer at 498 was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted July 11, 2025. ; https://doi.org/10.1101/2025.07.08.663620doi: bioRxiv preprint RT for 1 hour. After washed five times, samples were post-fixed with incubated with 4% 499 paraformaldehyde in 1× PBS for 10 minutes at RT. Finally, samples were stained with 1 μ g/mL 500 Hoechst 33342 diluted in 1× PBS for 10 minutes at RT. Cells in 35 mm Petri dishes were kept away 501 from light in 1× PBS at 4 /i2 until imaging. Cells on coverslips were mounted with Fluoromount-G and 502 dried in the shade, and kept away from light at 4 /i2 until imaging. Images were acquired using TCS 503 SP8 STED 3X (Leica) microscope with a ×100 objective and post-processed using Fiji (ImageJ) and 504 MA TLAB. 505 The heterogeneous index was calculated to reflect the uniformity of distribution like histone 506 modifications. Before batch processing, nucleoli and extranuclear regions were manually selected and 507 the intensity of these parts were set to 0. Then average intensity of each image was set to same value to 508 eliminate variations caused by sample preparation and imaging. Considering the shape of the aggregate 509 structures which form around 5-pixel thick sticks, we used a 5×5-pixel box traversing the image and 510 calculated the sum of intensity inside the box. The values of 0 were left out, and the histogram of 511 frequency distribution was calculated and fitted with double Gaussian distribution. The two Gaussian 512 curves represented the distribution of sparse regions and dense regions respectively. Distance between 513 the centers of two Gaussian peaks was calculated to quantify the aggregation level. It should be noted 514 that the rationality of fitting results is closely related to the starting point of fitting. Therefore, the 515 position with the highest frequency was used as the default starting point of the double Gaussian fitting, 516 and the starting point was adjusted manually for the results obviously not reasonable, such as negative 517 Gaussian curve strength. 518 Fluorescence in situ hybridization (FISH) 519 Cells used for FISH labeling were cultured on coverslips in 12-well plates. The whole experiment 520 procedure was according to Metasystems’ protocols. After washed with pre-warmed 1× PBS, cells were 521 fixed in pre-cooled 3:1 methanol/acetic-acid for 10 minutes at -20 /i2 . The fix reagent was removed and 522 the coverslip was air-dried. FISH probe targeting to chromosome 2 and 18 (Metasystems, 523 D-0302-100-FI XCP 2 and D-0318-100-OR XCP 18) were mixed with 5 μ L each and dropped on a 524 glass slide. The coverslip with cells was inverted onto the probes and sealed with nail polish. Then the 525 sample and probes were heated at 75 /i2 for 2 minutes to be denatured. Next, the sample was incubated 526 in a humidified chamber at 37 °C overnight. The next day, the coverslip was removed and washed in 527 0.4× SSC buffer (pH 7.0) at 72 °C for 2 minutes. Then the sample was washed in 2× SSC buffer 528 supplemented with 0.05% Tween-20 at RT for 30 seconds. Finally, the sample was stained with 1 529 μ g/mL Hoechst 33342 diluted in 1× PBS for 10 minutes at RT, mounted with Fluoromount-G, dried in 530 the shade, and kept away from light at 4 /i2 until imaging. Images were acquired using TCS SP8 STED 531 3X (Leica) microscope with a ×100 objective and post-processed using Fiji (ImageJ). 532 DNase I /Mnase digestion assays 533 Cells (approximately 2 ×/i2 106) were collected by centrifugation at 200 g for 3 minutes. After washed 534 by cold 1× PBS twice, cells were suspended and lysed by nucleus extraction buffer (Tris-HCl 10 mM, 535 pH 7.5, NaCl 10 mM, MgCl 2 3 mM, IGEPAL CA-630 0.1% and 0.1 mg/mL PMSF) on ice for 60 536 seconds. Then centrifuge immediately at 500 g for 10 minutes at 4 /i2 to pellet nuclei and remove 537 supernatant completely. Extracted nuclei were resuspended with 600 μ L DNase I reaction buffer 538 (Tris-HCl 10 mM, pH 7.5, MgCl 2 2.5 mM, CaCl 2 0.1 mM, sucrose 0.3M) or Mnase reaction buffer 539 (Tris-HCl 10 mM, pH 7.5, NaCl 10 mM, KCl 10 mM, MgCl2 3 mM, CaCl2 10 mM, sucrose 0.3M) and 540 was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted July 11, 2025. ; https://doi.org/10.1101/2025.07.08.663620doi: bioRxiv preprint divided into six tubes. Each tube was treated with different units of DNaseI (Thermo Fisher, EN0523) 541 or Mnase (Thermo Fisher, EN0181) at 37 /i2 and 800 rpm rotation for 10 minutes. The reactions were 542 terminated by 50 μ L 0.5 mM EDTA (pH 8.0) and 50 μ L 10% SDS. Then, equal volume of 2× DNA 543 extraction buffer (Tris-HCl 20 mM, pH 8.0, NaCl 200 mM, EDTA 100 mM, SDS 2%) was added to 544 each tube and samples were digested with 20 μ g/mL RnaseA (TransGene, GE101) at 37 /i2 for 30 545 minutes and 0.2 mg/mL Proteinase K at 65 /i2 for 1 hour. Finally, DNA was extracted with equal 546 volume of phenol/chloroform. Equal amounts of DNA (200 ng) were resolved on 1% agarose gel and 547 stained with EB. High MW chromatin was measured by ImageJ. 548 In situ Hi-C 549 The in situ Hi-C libraries were prepared and sequenced as the published method30. Briefly, cells were 550 grown to approximately 70–80% confluency, washed with 1×PBS, crosslinked using 1% formaldehyde, 551 and suspended in Hi-C lysis buffer (Tris-HCl 10 mM, pH8.0, NaCl 10mM, IGEPAL CA-630 0.2%) 552 supplemented with protease inhibitors. 100 Units DpnII restriction enzyme was added per sample at 553 37 /i2 with rotation for overnight chromatin digestion. DNA ends were marked with biotin and then 554 ligated together in situ . After crosslink reversal, the DNA was sheared to 300~500 bp fragments, and 555 then biotinylated ligation junctions were recovered with streptavidin beads. Hi-C libraries were 556 amplified using PCR. After that, sequencing adaptor was added. All sequence data were produced using 557 Illumina HiSeq X Ten paired-end sequencing. 558 Raw reads were firstly cut adaptor and filtered to generate cleaned fastq files using TrimGalore 559 (https://www.bioinformatics.babraham.ac.uk/projects/trim_galore/). Next, Hi-C data preprocessing was 560 done using HiC-Pro 64. Briefly, reads were first aligned on the hg19 reference genome. Uniquely 561 mapped reads were normalized using Iterative Correction and Eigenvector (ICE) decomposition and 562 library size. 563 For compartment A/B analysis, HiTC 65 was used to calculate the PC1 (at 150-kb resolution) with 564 custom R script. The whole genome was classified into two compartments, A and B compartment, 565 based on the positive or negative PC1 values. The compartment with higher gene density while the B 566 compartment with lower gene density. To investigated compartment switching, we compared the PC1 567 values between control and NuMA depleted HCT116 cells, using zero as the PC1 cutoff. We defined 568 the delta PC1 to describe A-to-B or B-to-A compartment switching, compaction and decompaction: 569 /g1856/g1857/g1864/g1872/g1853 /g1842/g18291 /g3404 /g1842/g18291 /g4666 /g1840/g1873/g1839/g1827 /g1856/g1857/g1868/g1864/g1857/g1872/g1861/g1867/g1866 /g4667 /g3398/g1842 /g1829 1 /g4666 /g1829/g1867/g1866/g1872/g1870/g1867/g1864 /g4667 If one compartment had the delta PC1 ≥ 0.005 and the sign of PC1 (NuMA-depletion) and PC1 570 (Control) were opposite, it would be defined as B-to-A compartment switching; if one compartment 571 had the delta PC1 ≤ -0.005 and the sign of PC1 (NuMA-depletion) and PC1 (Control) were opposite, it 572 would be defined as A-to-B compartment switching; if one compartment had the delta PC1 ≥ 0.005 and 573 the sign of PC1 (NuMA-depletion) and PC1 (Control) were identical, it would be defined as 574 compartment decompaction; if one compartment had the delta PC1 ≤ -0.005 and the sign of PC1 575 (NuMA-depletion) and PC1 (Control) were identical, it would be defined as compartment compaction. 576 Juicebox (http://aidenlab.org/juicebox/) was used to visualize Hi-C contact matrix 66. IGV 577 (https://igv.org) was used to display compartment switching. 578 was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted July 11, 2025. ; https://doi.org/10.1101/2025.07.08.663620doi: bioRxiv preprint ATAC-seq 579 Each sample (approximately 5 /i2 ×/i2 105 cells) was centrifuged with 500 g, 4 °C for 5 minutes to 580 collect cell pellet. The cell pellet was resuspended with moderately cold lysis buffer, and library 581 construction used the TruePrep DNA Library Prep Kit V2 for Illumina (V azyme, TD501). DNA 582 libraries were produced after 12 cycles of PCR amplification using the TruePrep library prep kit. After 583 purification, the paired-end sequencing was performed on the Illumina Novaseq 6000 platform. 584 Sequencing adaptors were removed from the raw A TAC-seq reads using TrimGalore 585 (https://www.bioinformatics.babraham.ac.uk/projects/trim_galore/), and the clean data were mapped to 586 the human reference genome (hg19) using Bowtie2 (v.2.4.1) 67. Picard 587 (https://broadinstitute.github.io/picard/) was used to remove PCR duplicates. 588 The nuclear repeat length (NRL) was calculated using NRLfinder as previous publication 33. Briefly, 589 read lengths were extracted and converted into a frequency histogram, which was then smoothed using 590 a digital 6th-order Butterworth filter with a zero-phase shift and a cutoff frequency of 0.04 cycles/read. 591 This cutoff was empirically optimized to reduce noise from mononucleosomal DNA winding artifacts. 592 Local minima and maxima were identified from the first derivative of the filtered histogram, with the 593 second peak maximum corresponding to the dinucleosomal periodicity. The NRL shift between 594 conditions (e.g., control vs. NuMA-depleted HCT116 cells) was calculated the mean difference 595 between the first two peak maxima of each sample. All analyses were performed in Python 3.9 with 596 NumPy, SciPy, and Matplotlib libraries. 597 For chromatin-state modelling, we used the ChromHMM (v.1.19)32. The input data of A TAC-seq and 598 RNA-seq reported in this manuscript was generated as described above. Additional input data including 599 ChIP-seq for CTCF, H3K4me3, H3K27me3, H3K4me1, H3K36me3 and H3K9me3 were download 600 from ENCODE (https://www.encodeproject.org). briefly, raw bam files were download and replicates 601 were combined. BinarizeBam and LearnModel tools in ChromHMM was used to generate chromatin 602 state model with default settings. Emissions parameters were visualized in R. 603 For LAD annotation of A TAC-seq peaks, the public data of LaminB1 DamID of HCT116 was 604 download from 4D Nucleome Data Portal (https://data.4dnucleome.org/) and analyzed as previous 605 publication9. 606 CUT&Tag 607 The CUT&Tag assay was performed using the NovoNGS® CUT&Tag 3.0 High-Sensitivity Kit 608 (NovoProtein, N259-YH01). 1/i2 ×/i2 105 cells were washed twice with 1.5 /i2 mL of wash buffer and then 609 mixed with activated concanavalin A beads. Primary antibody targeting histone H1 was added to the 610 system for 2 h at room temperature. After successive incubations with the primary antibody and 611 secondary antibody at RT for 1.5 hours. The cells were washed and incubated with pAG-Tn5 for 1.5/i2 h. 612 Then, tagmentation buffer was added to activate tagmentation for 1 /i2 h. The tagment DNA fragments 613 were purified using tagment DNA extract beads and washed by 80% ethanol. DNA libraries were 614 produced after PCR amplification and sequenced using the Illumina NovaSeq 6000 platform. 615 TrimGalore (https://www.bioinformatics.babraham.ac.uk/projects/trim_galore/) was used to cut 616 adaptors, and trimmed reads were aligned to the human genome (hg19) using Bowtie2 as described in 617 A TAC-seq data analysis. Reads were sorted and converted to BAM format and duplicates were marked 618 was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted July 11, 2025. ; https://doi.org/10.1101/2025.07.08.663620doi: bioRxiv preprint using Picard ( https://broadinstitute.github.io/picard/) using the MarkDuplicates module. After 619 normalizing the samples to the same sequencing depth, deepTools2 (v 3.3.1) software 68 was used to 620 plot the heat maps to show signals around peak regions with default parameters. Peaks were called 621 using Macs2 with a relaxed q value threshold of 0.001. 622 For peak calling, MACS2 69 (v2.2.7.1) was employed to identify enriched regions with the q value 623 threshold of 0.001 . Blacklisted regions (from the ENCODE Project) were excluded using bedtools 70 624 (v2.30.0). Peak regions were then normalized to the same sequencing depth using deepTools2 68 625 (v3.3.1), and heat maps of signal intensity around peak summits were generated with the 626 computeMatrix and plotHeatmap functions, setting the region of interest to ±3 kb from peak centers. 627 To analyze co-localization with histone modifications, peak intersections were performed using 628 bedtools intersect with a minimum overlap of 1 bp. Genomic annotation of peaks was conducted using 629 HOMER (v4.11.1). For visualization, IGV 71 was used to generate genome browser tracks, and 630 statistical analyses were performed in R with custom scripts. 631 RNA-seq 632 Total RNA was extracted using the MolPure Cell RNA Kit (YEASEN, 19231ES50). RNA 633 sequencing libraries were constructed using the NEBNext Ultra RNA Library Prep Kit for Illumina 634 (NEB, E7530). RNA-seq paired-end reads were sequenced on the Illumina NovaSeq 6000 platform. 635 The raw RNA sequences were cleaned using TrimGalore 636 (https://www.bioinformatics.babraham.ac.uk/projects/trim_galore/) and mapped to human reference 637 genome hg19 by STAR (v2.7.1a) with default parameters. All mapped bam files were converted to 638 bigwig using bedtools (v2.24.0)70 for visualization in IGV (https://igv.org). High-quality mapped reads 639 were quantified using htseq-count (v0.11.3) 72. Differentially expressed genes were analyzed by 640 DEseq273. GO enrichment analysis was performed using Enrichr74. The integrated analysis of RNA-seq 641 data and Hi-C data was done in custom R scripts. 642 Super-resolution imaging and live-cell imaging 643 STED imaging was performed using a commercial STED microscope (TCS SP8 STED 3X, Leica 644 Microsystems, Germany) equipped with an HCX PL APO ×100/1.40 NA objective. Alexa 594 labeled 645 samples were excited with white laser at the wavelength of 594 nm, and Atto 674N labeled samples 646 were excited with white laser at the wavelength of 647 nm. STED laser for Alexa 594 was continuous 647 wave at the wavelength of 660 nm, and the wavelength of STED laser for Atto 674N was 775 nm 648 pulsed laser. All images were acquired with the LAS AF software (Leica). De-convolution was 649 performed with LAS AF software (Leica) and post-processed using Fiji (ImageJ). 650 HeLa cells co-transfected with the eGFP-H1.1 and different NuMA truncations fused with mScarlet 651 were passaged into 35 mm Petri dishes to 60-70 % confluency. Imaging was performed using a 652 spinning disk confocal system (Live SR CSU W Nikon) with an EMCCD (iXon DU-897E) mounted on 653 a Nikon Ti-E microscope with a CFI Apo TIRF ×100 Oil (N.A. 1.49) objective. During image 654 acquisition, the cells were incubated at 37 °C with 5% CO 2. All images were post-processed using Fiji 655 (ImageJ). 656 Expansion sample preparation and image acquisition 657 To prepare hydrogel samples, we performed the gelation processes referring to Joerg Bewersdorf and 658 was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted July 11, 2025. ; https://doi.org/10.1101/2025.07.08.663620doi: bioRxiv preprint the co-workers’ instructions 41, 42 . Before gelation, fixed cells were labeled with first antibody and 659 second antibody as described above. Then the samples were fixed with EM-grade 3% PFA (Electron 660 Microscopy Sciences, 157-8) and 0.1% glutaraldehyde (GA, Electron Microscopy Sciences, 16020) in 661 1× PBS for 15 min at RT with rotation. After washed with 1× PBS for 3 times, cells were post-fixed in 662 1% acrylamide (AAm, Sigma-Aldrich, A9099) and 0.7% PFA at 37ºC with shaking for 4 hours and 663 washed with 1× PBS twice for 10 min at RT. Then samples were rinsed with gelling solution consisted 664 of 10% AAm, 0.1% N, N'-(1,2-dihydroxyethylene) bisacrylamide (DHEBA, Sigma-Aldrich, 294381), 665 and 19% sodium acrylate (SA, Sigma-Aldrich, 408220) and 1× PBS. 0.01% ammonium persulfate 666 (APS, Sigma-Aldrich, A3678) and 0.01% N, N, N’, N’-tetramethylethylenediamine (TEMED, 667 Sigma-Aldrich, T22500) were added immediately before use to activate the gelling solution. Activated 668 gelling solution was applied to samples, incubated at RT for 15 min and at 37 ºC for 1 hour. Then the 669 hydrogel was isolated and incubated in denaturation buffer (50 mM Tris-HCl, pH 6.8, 200 mM NaCl, 670 200 mM SDS) for 1 hour at 76 ºC with rotation. After denaturation, the hydrogel was transferred to 671 ultrapure water for two 30 min washes, and then left to expand overnight. Before image acquisition, the 672 hydrogel was incubated in ultrapure water containing 1 μ g/mL Hoechst 33342 at RT for 15 min. Then 673 the hydrogel was put on 40 mm glass bottom dish coated with poly-L-lysine and imaged. 674 Protein purification 675 Plasmids of pET28a-6×His-mScarlet-NuMA, pET28a-GST-NuMA and NuMA ’s truncations were 676 transferred into BL21(DE3) pLysS chemically competent cell (TransGene, CD701). Single colony was 677 picked and cultured in Luria–Bertani (LB) medium supplemented with 30 mg/mL kanamycin at 37 °C 678 overnight with shaking at 220 rpm. The culture was inoculated into fresh LB medium at the ratio of 679 1:50 to expand the culture to OD 0.6, then cooled to 18 °C. 0.5 mM IPTG was added for approximately 680 18 hours to induce protein expression. Cells were harvested by centrifugation at 5000 rpm, for 20 681 minutes 4 °C and resuspended with resuspension buffer (Tris-HCl 50 mM, pH 7.5, NaCl 2 M, glycerol 682 10%, 0.1 mg/mL PMSF, and 1 mM DTT for GST-tagged proteins). All purification steps were 683 performed on ice or at 4 °C to maintain protein activity. Cell suspension was sonicated on ice, and the 684 lysates were treated with 0.25 % PEI to eliminate DNA contamination and centrifuged at 18,000 g for 685 20 minutes at 4 °C. The supernatant was incubated with Ni-NTA (for proteins with 6×His tag) or GST 686 agarose resin (for proteins with GST tag) at 4 °C for 1 hour. Ni-NTA beads were washed extensively 687 with His washing buffer (Tris-HCl 50 mM, pH 7.5, 500 mM NaCl, imidazole 50 mM, glycerol 10%), 688 and recombinant proteins were eluted with His elution buffer (Tris-HCl 50 mM, pH 7.5, 300 mM NaCl, 689 imidazole 200 mM, glycerol 10%). GST agarose beads were washed extensively with GST washing 690 buffer (Tris-HCl 50 mM, pH 7.5, 500 mM NaCl, glycerol 10%, 1 mM DTT) and recombinant proteins 691 were eluted with GST washing buffer supplemented with 20 mM GSH. 692 For purification of histone H1’s isoforms, similar procedures were performed as 6×His-tagged 693 NuMA. Plasmids of pET28a-6×His-HA-H1 were transferred into BL21(DE3) pLysS chemically 694 competent cell and the protein expression was induced by 0.5 mM IPTG at 18 °C for approximately 18 695 hours. The protein was purified using Ni-NTA followed by CM Sepharose. The CM beads were 696 extensively washed by GST washing buffer and histone H1 was eluted with CM elution buffer 697 (Tris-HCl 50 mM, pH 7.5, 800 mM NaCl, glycerol 10%). All eluted fractions were analyzed by 698 SDS-PAGE and Coomassie staining. Purified proteins were concentrated by ultra-centrifugation and 699 exchanged via dialysis into storage buffer (Tris-HCl 50 mM, pH 7.5, 150 mM NaCl), flash-frozen in 700 liquid nitrogen, sub-packaged and stored at -80 °C. 701 was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted July 11, 2025. ; https://doi.org/10.1101/2025.07.08.663620doi: bioRxiv preprint Co-IP and pull-down 702 For immunoprecipitation, HEK293 cells transfected with expression plasmids or HCT116 cells with 703 affinity tags knocked-in at NuMA gene locus were harvested by centrifugation at 200 g, 4 °C for 3 704 minutes. Cell pellets were directly lysed with IP lysis buffer (Tris-HCl 10 mM, pH 7.5, NaCl 150 mM, 705 TritonX-100 1%, sodium deoxycholate 1%, EDTA 1mM) supplemented with 0.1 mg/mL PMSF, 706 complete protease inhibitor and UltraNuclease on ice for 15 minutes. After lysis, the lysates were 707 centrifuged for 10 minutes at 21,100 g and 4 °C. 1% of the supernatant was stored as input and the 708 other was incubated with HA or streptavidin beads and shaking at 4 °C overnight. The next day, beads 709 were washed by IP washing buffer (Tris-HCl 10 mM, pH 7.5, NaCl 500 mM, Tween-20 0.5%) for five 710 times. Finally, beads with immunoprecipitated proteins were eluted with SDS-PAGE loading buffer and 711 analyzed with western-blotting. 712 Purified GST or GST-NuMA and NuMA ’s truncations recombinant proteins were bound to 713 GST-magnetic beads for 2 hours at 4°C in binding buffer (Tris-HCl 10 mM, pH 7.5, NaCl 300 mM, 714 Tween-20 0.1%, 10% glycerol), washed once with binding buffer and incubated with purified 715 HA-tagged histone H1 overnight at 4°C with slow rotation. The next day, beads were washed three five 716 with washing buffer (Tris-HCl 10 mM, pH 7.5, NaCl 500 mM, Tween-20 0.5%, 10% glycerol). Finally, 717 pull-downed proteins were eluted with elution buffer (Glycine 0.1M, pH 2.5) and added with 718 SDS-PAGE loading buffer supplemented with 50 mM Tris-HCl, pH8.8. Then samples were analyzed 719 with western-blotting and SDS-PAGE. 720 For pulldown of NuMA ’s C-terminal domain using immobilized nucleosomes, the 721 mono-nucleosomes and NuMA-C were incubated with streptavidin agarose beads with or without 722 histone H1 in the binding buffer (Tris–HCl 10 /i2 mM, pH 7.5, NaCl 150 mM, EDTA 0.2 mM, glycerol 723 10%, NP-40 0.1%) at 4 /i2 °C overnight. After incubation, the bound proteins were washed five times 724 with the washing buffer (Tris–HCl 10 /i2 mM, pH 7.5, NaCl 300 mM, EDTA 0.2 mM, glycerol 10%, 725 NP-40 0.5%, PMSF 0.5/i2 mM) and were analyzed by Western blotting. 726 Electrophoretic mobility shift assay (EMSA) 727 The DNA substrates used in this study were 5’-HEX labeled and resolved in reaction buffer 728 (Tris-HCl 10 mM, pH 7.5, NaCl 150 mM). DNA substrates was commercially ordered single-strands 729 were premixed and annealed to double-strand substrate. The DNA sequences are listed in 730 Supplemented Table 2. Purified Histone H1 and NuMA ’s truncations were mixed with dsDNA 731 substrates and incubate at RT for 15 minutes. Then samples were added with 5% glycerol and loaded to 732 1% agarose gel and run electrophoresis at 80V for 1 hour in 1×TBE buffer (Tris-HCl 90 mM, pH 8.0, 733 Boric acid 90 mM， EDTA 2 mM). HEX labeled DNA was visualized using Bio-Rad GelDOC XR. 734 Nucleosome assembly and sucrose density gradient centrifugation 735 Mono-nucleosome, 4× and 12× nucleosome array (NA) were assembled as previously described 40. 736 Binding of NuMA ’s C-terminal domain with H1 and 12× NA at a ratio of 1:2 was incubated in 737 centrifugation buffer (HEPES 10 mM, pH 8.0, EDTA 0.2 mM, 25 mM NaCl) at RT for 30 minutes. The 738 sucrose density gradient was set to 10-30 % using Gradient Master Model (Biocomp, 108). The 739 centrifugation was performed on Beckman Coulter XL-I using a sw-60Ti rotor at 40,000× g and 4 °C 740 under a vacuum for 16 hours. After the centrifugation was accomplished, the samples were divided into 741 was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted July 11, 2025. ; https://doi.org/10.1101/2025.07.08.663620doi: bioRxiv preprint 17 fractions at same volume and then resolved on 1% agarose gel which was stained with EB (Thermo 742 Fisher, 15585011). 743 Atomic force microscopy (AFM) 744 NuMA ’s C-terminal domain and histone H1 were incubated with 12× NA at RT for 15 minutes in 745 AFM buffer (HEPES 20 mM, pH 7.5, 150 mM NaCl) then fixed with 0.1% glutaraldehyde on ice for 746 30 min. During the incubation, 20 μ L spermidine (1 mM, Sigma-Aldrich, S2626) was dropped onto 747 newly cleaved mica surface and incubate for 10 min at RT. Then rinse the mica with 200 μ L ddH2O for 748 4 times and blow dry mica surface briefly. Next, the samples were diluted to 1 ng/μ L and 10 μ L sample 749 solution were added onto mica surface, incubate for 10 min. Finally, wash the mica with 200 μ L ddH2O 750 for 3 times and blow dry gently. The prepared AFM samples were examined using Bio Atomic Force 751 Microscope (BioScope Resolve, Bruker) and images were post-processed using Nanoscope Analysis 752 (Bruker). 753 Cryo-focused ion beam (FIB) milling and cryo-electron tomography data 754 collection and processing 755 U2OS cells were seeded on grids (QUANTIFOIL SiO2) in DMEM medium 8 hours after transfection. 756 Excess medium was removed by manually blotting with filter paper from the back of the grids. 4 µL of 757 cryo buffer (cell medium with 8% v/v final concentration glycerol) was added to the grid before 758 plunging to avoid nucleus unvitrification. Grids were blotted using a Vitrobot Mark IV at 90% 759 humidity and 37 °C, with a blot force of 10 and a blot time of 10 seconds. Cells were plunge-frozen in 760 an ethane/propane mixture cooled to liquid nitrogen temperatures. Grids were clipped into homemade 761 cut-off autogrid support rings and stored in liquid nitrogen to facilitate downstream handling. Grids 762 were loaded into an Aquilos2 cryo-focused ion beam/scanning electron microscope (FIB/SEM 763 dual-beam microscope, Thermo Fisher Scientific) maintained at liquid nitrogen temperature throughout 764 the procedure. To improve sample conductivity and reduce curtaining artifacts during FIB milling, 765 grids were first sputter-coated with platinum and then coated with organometallic platinum using the 766 gas injection system. More than 10 target cells from each strain were chosen randomly. Lamellae were 767 prepared using a gallium ion beam at 30 kV and stage tilt angles of 13°. Rough milling was performed 768 with currents of 0.3 nA until lamellae thickness reached ~800 nm. Fine milling of the lamellae was 769 gradually reduced to lower currents, with 30 pA used for the final polishing step. 770 Tilt series were collected using a Titan Krios G4 instrument at 300 kV , equipped with a Selectris X 771 energy filter and a Falcon 4i camera (Thermo Fisher Scientific). Tilt series were recorded using the 772 Tomo5 software package (Thermo Fisher Scientific). A dose-symmetric tilt scheme was used with an 773 angular increment of 3°. Magnification was set to 64,000×, with a pixel size of 1.89 Å and a dose of 774 3.5 e−/Ų per tilt, for a total dose of ~140 e-/Ų. The target defocus ranged from -4.0 to -5.0 µm. Data 775 were processed using the TOMOMAN 75 package version 0.7 pipeline 776 (https://github.com/williamnwan/TOMOMAN). After frame motion correction, bad tilts were removed 777 after manual inspection using a TOMOMAN script. Tilt series were split into odd and even tilts during 778 motion correction for denoising, and the resulting stacks were denoised using CryoCARE 76. Tilt series 779 were aligned using AreTomo77 version 1.3.3. Tomogram reconstructions were performed with IMOD78 780 version 4.12.32 at bin4 for visualization and bin2 for nucleosome template matching. Nucleosome 781 coordinates were determined using the template matching routine from STOPGAP 79 version 0.7. A 782

nucleosome core particle density map (EMD-8140)80 was lowpass filtered to 30 Å to serve as 783 was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted July 11, 2025. ; https://doi.org/10.1101/2025.07.08.663620doi: bioRxiv preprint the template. Subtomograms were binned by a factor of 2 (pixel size 3.78 Å/pixel), extracted by Warp81 784 from best quality 15 tomograms (10 tomograms from NuMA-dC, 5 tomograms from NuMA-FL). After 785 3D classification in RELION 3.0 82 with spherical mask, 4709 particles from NuMA-dC and 3601 786 particles from NuMA-FL reasonable classes were retained for auto-refinement. The half maps and 787 particle lists were imported into M 83 for geometric refinement. Bin1 (pixel size 1.89 Å/pixel) 788 subtomograms were re-extracted after M83 refinement and imported into RELION82 for a second round 789 of 3D classification. Relative particles were retained for M geometric and CTF parameter refinement. 790 After reaching resolution convergence, a final map at 9.9 Å resolution from dC cells and 17.8 Å from 791 FL cells were obtained within the reconstruction mask using the 0.143 FSC criterion. 792 NND analysis was based on nucleosome coordinates after subtomogram alignment. Nucleosome 793 core particle (PDB:2CV5)84 and chromatosome (PDB: 7PF6)85 were applied for EM map docking. 794 Data availability 795 The sequencing data of Hi-C, A TAC-seq, RNA-seq and Cut&Tag analyzed in this study are 796 accessible at the NCBI GEO website under the accession number GSE227941. R markdown scripts 797 enabling the main analyses are available from the responding author upon reasonable request. EM 798 density maps have been deposited to EMDB under codes EMD-51858, EMD-51859. All other data 799 supporting the findings of this study are available from the corresponding author on reasonable request. 800 Code availability 801 Code use in this article can be made available upon request to the corresponding author. 802

803 1. Grewal, S.I. & Jia, S. Heterochromatin revisited. Nat Rev Genet 8, 35-46 (2007). 804 2. Allshire, R.C. & Madhani, H.D. Ten principles of heterochromatin formation and function. 805 Nat Rev Mol Cell Biol 19, 229-244 (2018). 806 3. Janssen, A., Colmenares, S.U. & Karpen, G.H. Heterochromatin: Guardian of the Genome. 807 Annu Rev Cell Dev Biol 34, 265-288 (2018). 808 4. Greenstein, R.A. & Al-Sady, B. Epigenetic fates of gene silencing established by 809 heterochromatin spreading in cell identity and genome stability. Curr Genet 65, 423-428 810 (2019). 811 5. Grewal, S.I.S. The molecular basis of heterochromatin assembly and epigenetic inheritance. 812 Mol Cell 83, 1767-1785 (2023). 813 6. Fan, H. et al. The nuclear matrix protein HNRNPU maintains 3D genome architecture 814 globally in mouse hepatocytes. Genome Res 28, 192-202 (2018). 815 7. Nozawa, R.S. et al. SAF-A Regulates Interphase Chromosome Structure through 816 Oligomerization with Chromatin-Associated RNAs. Cell 169, 1214-1227 e1218 (2017). 817 was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted July 11, 2025. ; https://doi.org/10.1101/2025.07.08.663620doi: bioRxiv preprint 8. Ma, G. et al. The nuclear matrix stabilizes primed-specific genes in human pluripotent stem 818 cells. Nat Cell Biol 27, 232-245 (2025). 819 9. Chang, L. et al. Nuclear peripheral chromatin-lamin B1 interaction is required for global 820 integrity of chromatin architecture and dynamics in human cells. Protein Cell 13, 258-280 821 (2020). 822 10. Huo, X. et al. The Nuclear Matrix Protein SAFB Cooperates with Major Satellite RNAs to 823 Stabilize Heterochromatin Architecture Partially through Phase Separation. Mol Cell 77, 824 368-383 e367 (2020). 825 11. Zhang, Y . et al. MA TR3-antisense LINE1 RNA meshwork scaffolds higher-order chromatin 826 organization. EMBO Rep 24, e57550 (2023). 827 12. Berezney, R. & Coffey, D.S. Nuclear matrix. Isolation and characterization of a framework 828 structure from rat liver nuclei. J Cell Biol 73, 616-637 (1977). 829 13. Harborth, J., Wang, J., Gueth-Hallonet, C., Weber, K. & Osborn, M. Self assembly of NuMA: 830 multiarm oligomers as structural units of a nuclear lattice. EMBO J 18, 1689-1700 (1999). 831 14. Zeng, C., He, D. & Brinkley, B.R. Localization of NuMA protein isoforms in the nuclear 832 matrix of mammalian cells. Cell Motil Cytoskeleton 29, 167-176 (1994). 833 15. Engelke, R. et al. The quantitative nuclear matrix proteome as a biochemical snapshot of 834 nuclear organization. J Proteome Res 13, 3940-3956 (2014). 835 16. Lydersen, B.K. & Pettijohn, D.E. Human-specific nuclear protein that associates with the 836 polar region of the mitotic apparatus: distribution in a human/hamster hybrid cell. Cell 22, 837 489-499 (1980). 838 17. Compton, D.A., Szilak, I. & Cleveland, D.W. Primary Structure of NuMA, an Intranuclear 839 Protein That Defines a Novel Pathway for Segregation of Proteins at Mitosis. Journal of Cell 840 Biology 116, 1395-1408 (1992). 841 18. Kempf, T., Bischoff, F.R., Kalies, I. & Ponstingl, H. Isolation of Human NuMA Protein. Febs 842 Lett 35 4 , 307-310 (1994). 843 19. Radulescu, A.E. & Cleveland, D.W. NuMA after 30 years: the matrix revisited. Trends Cell 844 Biol 20, 214-222 (2010). 845 20. Kiyomitsu, T. & Boerner, S. The Nuclear Mitotic Apparatus (NuMA) Protein: A Key Player 846 for Nuclear Formation, Spindle Assembly, and Spindle Positioning. Front Cell Dev Biol 9, 847 653801 (2021). 848 21. Luderus, M.E., den Blaauwen, J.L., de Smit, O.J., Compton, D.A. & van Driel, R. Binding of 849 matrix attachment regions to lamin polymers involves single-stranded regions and the minor 850 groove. Mol Cell Biol 14, 6297-6305 (1994). 851 22. Rajeevan, A., Keshri, R., Kapoor, S. & Kotak, S. NuMA interaction with chromatin is vital for 852 proper chromosome decondensation at the mitotic exit. Mol Biol Cell 31, 2437-2451 (2020). 853 23. Serra-Marques, A. et al. The mitotic protein NuMA plays a spindle-independent role in 854 was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted July 11, 2025. ; https://doi.org/10.1101/2025.07.08.663620doi: bioRxiv preprint nuclear formation and mechanics. J Cell Biol 219, 1-18 (2020). 855 24. Kivinen, K., Taimen, P. & Kallajoki, M. Silencing of Nuclear Mitotic Apparatus protein 856 (NuMA) accelerates the apoptotic disintegration of the nucleus. Apoptosis 15, 936-945 (2010). 857 25. Abad, P.C. et al. NuMA influences higher order chromatin organization in human mammary 858 epithelium. Mol Biol Cell 18, 348-361 (2007). 859 26. Gueth-Hallonet, C., Wang, J., Harborth, J., Weber, K. & Osborn, M. Induction of a regular 860 nuclear lattice by overexpression of NuMA. Exp Cell Res 243, 434-452 (1998). 861 27. Hergeth, S.P. & Schneider, R. The H1 linker histones: multifunctional proteins beyond the 862 nucleosomal core particle. EMBO Rep 16, 1439-1453 (2015). 863 28. Natsume, T., Kiyomitsu, T., Saga, Y . & Kanemaki, M.T. Rapid Protein Depletion in Human 864 Cells by Auxin-Inducible Degron Tagging with Short Homology Donors. Cell Rep 15, 865 210-218 (2016). 866 29. Silk, A.D., Holland, A.J. & Cleveland, D.W. Requirements for NuMA in maintenance and 867 establishment of mammalian spindle poles. J Cell Biol 184, 677-690 (2009). 868 30. Rao, S.S. et al. A 3D map of the human genome at kilobase resolution reveals principles of 869 chromatin looping. Cell 159, 1665-1680 (2014). 870 31. Buenrostro, J.D., Giresi, P.G., Zaba, L.C., Chang, H.Y . & Greenleaf, W.J. Transposition of 871 native chromatin for fast and sensitive epigenomic profiling of open chromatin, DNA-binding 872 proteins and nucleosome position. Nat Methods 10, 1213-1218 (2013). 873 32. Ernst, J. & Kellis, M. ChromHMM: automating chromatin-state discovery and 874 characterization. Nat Methods 9, 215-216 (2012). 875 33. Willcockson, M.A. et al. H1 histones control the epigenetic landscape by local chromatin 876 compaction. Nature 589, 293-298 (2021). 877 34. Ricci, M.A., Manzo, C., Garcia-Parajo, M.F., Lakadamyali, M. & Cosma, M.P. Chromatin 878 fibers are formed by heterogeneous groups of nucleosomes in vivo. Cell 160, 1 145-1158 879 (2015). 880 35. Fyodorov, D.V ., Zhou, B.R., Skoultchi, A.I. & Bai, Y . Emerging roles of linker histones in 881 regulating chromatin structure and function. Nat Rev Mol Cell Biol 19, 192-206 (2018). 882 36. Woodcock, C.L., Skoultchi, A.I. & Fan, Y . Role of linker histone in chromatin structure and 883 function: H1 stoichiometry and nucleosome repeat length. Chromosome Res 14, 17-25 (2006). 884 37. Baldi, S., Korber, P. & Becker, P.B. Beads on a string-nucleosome array arrangements and 885 folding of the chromatin fiber. Nat Struct Mol Biol 27, 109-118 (2020). 886 38. Kaya-Okur, H.S., Janssens, D.H., Henikoff, J.G., Ahmad, K. & Henikoff, S. Efficient low-cost 887 chromatin profiling with CUT&Tag. Nature Protocols 15, 3264-3283 (2020). 888 39. Hizume, K., Y oshimura, S.H. & Takeyasu, K. Linker histone H1 per se can induce 889 three-dimensional folding of chromatin fiber. Biochemistry 44, 12978-12989 (2005). 890 was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted July 11, 2025. ; https://doi.org/10.1101/2025.07.08.663620doi: bioRxiv preprint 40. Song, F. et al. Cryo-EM study of the chromatin fiber reveals a double helix twisted by 891 tetranucleosomal units. Science 344, 376-380 (2014). 892 41. Pownall, M.E. et al. Chromatin expansion microscopy reveals nanoscale organization of 893 transcription and chromatin. Science 381, 92-99 (2023). 894 42. M'Saad, O. & Bewersdorf, J. Light microscopy of proteins in their ultrastructural context. 895 Nature Communications 11 (2020). 896 43. Lu, X. et al. Drosophila H1 regulates the genetic activity of heterochromatin by recruitment of 897 Su(var)3-9. Science 340, 78-81 (2013). 898 44. Liu, C.F. et al. Histone H1 facilitates restoration of H3K27me3 during DNA replication by 899 chromatin compaction. Nature Communications 14 (2023). 900 45. Healton, S.E. et al. H1 linker histones silence repetitive elements by promoting both histone 901 H3K9 methylation and chromatin compaction. Proc Natl Acad Sci U S A 117, 14251-14258 902 (2020). 903 46. He, S. et al. Linker histone H1 drives heterochromatin condensation via phase separation in 904 Arabidopsis. Plant Cell (2024). 905 47. Lieberman-Aiden, E. et al. Comprehensive mapping of long-range interactions reveals folding 906 principles of the human genome. Science 326, 289-293 (2009). 907 48. Gibson, B.A. et al. Organization of Chromatin by Intrinsic and Regulated Phase Separation. 908 Cell 179, 470-484 e421 (2019). 909 49. Lu, X. et al. Linker histone H1 is essential for Drosophila development, the establishment of 910 pericentric heterochromatin, and a normal polytene chromosome structure. Genes Dev 23, 911 452-465 (2009). 912 50. Hayward, A. & Gilbert, C. Transposable elements. Current Biology 32, R 904-R909 (2022). 913 51. Wells, J.N. & Feschotte, C. A Field Guide to Eukaryotic Transposable Elements. Annual 914 Review of Genetics, V ol 54, 2020 54, 539-561 (2020). 915 52. Xu, R.M. et al. Stage-specific H3K9me3 occupancy ensures retrotransposon silencing in 916 human pre-implantation embryos. Cell Stem Cell 29, 1051-+ (2022). 917 53. Liu, X.Q. et al. Resurrection of endogenous retroviruses during aging reinforces senescence. 918 Cell 186, 287-+ (2023). 919 54. Yan, L. et al. Single-cell RNA-Seq profiling of human preimplantation embryos and 920 embryonic stem cells. Nat Struct Mol Biol 20, 1131-1139 (2013). 921 55. Wu, J. et al. Chromatin analysis in human early development reveals epigenetic transition 922 during ZGA. Nature 557, 256-260 (2018). 923 56. Deng, Q., Ramsköld, D., Reinius, B. & Sandberg, R. Single-cell RNA-seq reveals dynamic, 924 random monoallelic gene expression in mammalian cells. Science 343, 193-196 (2014). 925 57. Turner, A.L. et al. Highly disordered histone H1-DNA model complexes and their condensates. 926 was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted July 11, 2025. ; https://doi.org/10.1101/2025.07.08.663620doi: bioRxiv preprint Proc Natl Acad Sci U S A 115, 11964-11969 (2018). 927 58. Wang, L. et al. Rett syndrome-causing mutations compromise MeCP2-mediated liquid-liquid 928 phase separation of chromatin. Cell Res 30, 393-407 (2020). 929 59. Sun, M. et al. NuMA regulates mitotic spindle assembly, structural dynamics and function via 930 phase separation. Nature Communications 12 (2021). 931 60. Ma, H. et al. NuMA forms condensates through phase separation to drive spindle pole 932 assembly. Journal of Molecular Cell Biology 14 (2022). 933 61. Okumura, M., Natsume, T., Kanemaki, M.T. & Kiyomitsu, T. Dynein-Dynactin-NuMA 934 clusters generate cortical spindle-pulling forces as a multi-arm ensemble. Elife 7, 1-24 (2018). 935 62. Jackson, D.A. & Pombo, A. Replicon clusters are stable units of chromosome structure: 936 Evidence that nuclear organization contributes to the efficient activation and propagation of S 937 phase in human cells. Journal of Cell Biology 140, 1285-1295 (1998). 938 63. Dimitrova, D.S., Todorov, I.T., Melendy, T. & Gilbert, D.M. Mcm2, but not RPA, is a 939 component of the mammalian early G1-phase prereplication complex. J Cell Biol 146, 940 709-722 (1999). 941 64. Servant, N. et al. HiC-Pro: an optimized and flexible pipeline for Hi-C data processing. 942 Genome Biol. 16, 259 (2015). 943 65. Servant, N. et al. HiTC: exploration of high-throughput 'C' experiments. Bioinformatics 28, 944 2843-2844 (2012). 945 66. Durand, N.C. et al. Juicer Provides a One-Click System for Analyzing Loop-Resolution Hi-C 946 Experiments. Cell Syst 3, 95-98 (2016). 947 67. Langmead, B. & Salzberg, S.L. Fast gapped-read alignment with Bowtie 2. Nat Methods 9, 948 357-359 (2012). 949 68. Ramirez, F. et al. deepTools2: a next generation web server for deep-sequencing data analysis. 950 Nucleic Acids Res 44, W160-165 (2016). 951 69. Zhang, Y . et al. Model-based analysis of ChIP-Seq (MACS). Genome Biol 9, R137 (2008). 952 70. Quinlan, A.R. BEDTools: The Swiss-Army Tool for Genome Feature Analysis. Curr Protoc 953 Bioinformatics 47, 11 12 11-34 (2014). 954 71. Robinson, J.T., Thorvaldsdottir, H., Turner, D. & Mesirov, J.P. igv.js: an embeddable 955 JavaScript implementation of the Integrative Genomics Viewer (IGV). Bioinformatics 39 956 (2023). 957 72. Anders, S., Pyl, P.T. & Huber, W. HTSeq--a Python framework to work with high-throughput 958 sequencing data. Bioinformatics 31, 166-169 (2015). 959 73. Love, M.I., Huber, W. & Anders, S. Moderated estimation of fold change and dispersion for 960 RNA-seq data with DESeq2. Genome Biol 15, 550 (2014). 961 74. Kuleshov, M.V . et al. Enrichr: a comprehensive gene set enrichment analysis web server 2016 962 was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted July 11, 2025. ; https://doi.org/10.1101/2025.07.08.663620doi: bioRxiv preprint update. Nucleic Acids Res 44, W90-97 (2016). 963 75. Khavnekar, S., Erdmann, P. & Wan, W. TOMOMAN: Streamlining Cryo-electron 964 Tomography and Subtomogram Averaging Workflows Using TOMOgram MANager. 965 Microscopy and Microanalysis 29, 1020-1020 (2023). 966 76. Buchholz, T.-O., Jordan, M., Pigino, G. & Jug, F. in 2019 IEEE 16th International Symposium 967 on Biomedical Imaging (ISBI 2019) 502-506 (2019). 968 77. Zheng, S. et al. AreTomo: An integrated software package for automated marker-free, 969 motion-corrected cryo-electron tomographic alignment and reconstruction. Journal of 970 Structural Biology: X 6 (2022). 971 78. Kremer, J.R., Mastronarde, D.N. & McIntosh, J.R. Computer Visualization of 972 Three-Dimensional Image Data Using IMOD. Journal of Structural Biology 116, 71-76 973 (1996). 974 79. Wan, W., Khavnekar, S. & Wagner, J. STOPGAP: an open-source package for template 975 matching, subtomogram alignment and classification. Acta Crystallographica Section D 976 Structural Biology 80, 336-349 (2024). 977 80. Chua, E.Y .D. et al. 3.9 Å structure of the nucleosome core particle determined by phase-plate 978 cryo-EM. Nucleic Acids Research 44, 8013-8019 (2016). 979 81. Tegunov, D. & Cramer, P. Real-time cryo-electron microscopy data preprocessing with Warp. 980 Nat Meth 16, 1146-1152 (2019). 981 82. Zivanov, J. et al. New tools for automated high-resolution cryo-EM structure determination in 982 RELION-3. eLife 7 (2018). 983 83. Tegunov, D., Xue, L., Dienemann, C., Cramer, P. & Mahamid, J. Multi-particle cryo-EM 984 refinement with M visualizes ribosome-antibiotic complex at 3.5 /i2 Å in cells. Nat Meth 18, 985 186-193 (2021). 986 84. Tsunaka, Y . Alteration of the nucleosomal DNA path in the crystal structure of a human 987 nucleosome core particle. Nucleic Acids Research 33, 3424-3434 (2005). 988 85. Dombrowski, M., Engeholm, M., Dienemann, C., Dodonova, S. & Cramer, P. Histone H1 989 binding to nucleosome arrays depends on linker DNA length and trajectory. Nature Structural 990 & Molecular Biology 29, 493-501 (2022). 991 992 Acknowledgments 993 We thank Prof. Dongyi Xu for providing HCT116 cells. We thank National Center for Protein 994 Science at Peking University in Beijing, China, for assistance with flow cytometry and imaging, 995 particularly Ms. Liying Du, Dr. Chunyan Shan, Dr. Liqin Fu and Dr. Yiqun Liu for technical help. This 996 work is supported by grants from the National Key R&D Program of China No. 2022YFA1303103 and 997 2022YFA3401100, and the National Science Foundation of China 22127804 for Y .S. 998 was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted July 11, 2025. ; https://doi.org/10.1101/2025.07.08.663620doi: bioRxiv preprint Author information 999 These authors contributed equally: Yao Wang, Wenxue Zhao, Jiahao Niu. 1000 Authors and Affiliations 1001 National Biomedical Imaging Center, College of Future T echnology, Peking University, Beijing, 1002 China 1003 Yao Wang, Jiahao Niu, Xiaotian Wang, Weihong Y uan, Aibin He & Y ujie Sun 1004 State Key Laboratory of Membrane Biology & Biomedical Pioneering Innovation Center 1005 (BIOPIC), Peking University, Beijing, China 1006 Yao Wang, Jiahao Niu, Xiaotian Wang, Weihong Y uan & Y ujie Sun 1007 School of Life Sciences, Peking University, Beijing, China 1008 Wenxue Zhao, Xiaotian Wang, Weihong Y uan, Cheng Li & Y ujie Sun 1009 Center for Bioinformatics, Center for Statistical Science, Peking University, Beijing, China 1010 Wenxue Zhao & Cheng Li 1011 Key Laboratory of Epigenetic Regulation and Intervention, Chinese Academy of Sciences, 1012 Beijing, China 1013 Cuifang Liu & Guohong Li 1014 Guangdong Provincial Key Laboratory of Bone and Joint Degeneration Diseases, Department of 1015 Physiology, School of Basic Medical Sciences, Southern Medical University, Guangzhou, China. 1016 Shanshan Ai 1017 Department of Cardiology, Heart Center, Zhujiang Hospital, Southern Medical University, 1018 Guangzhou, China. 1019 Shanshan Ai 1020 Department of Molecular Structural Biology, Max Planck Institute of Biochemistry, Am 1021 Klopferspitz 18, Martinsried, Germany 1022 Wolfgang Baumeister & Peng Xu 1023 New Cornerstone Science Laboratory, Frontier Science Center for Immunology and Metabolism, 1024 Hubei Key Laboratory of Cell Homeostasis, College of Life Sciences, Taikang Center for Life and 1025 Medical Sciences, Wuhan University, Wuhan, China 1026 Guohong Li 1027 Institute of Molecular Medicine, Peking-Tsinghua Center for Life Sciences, Peking University, 1028 Beijing, China 1029 Aibin He 1030 was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted July 11, 2025. ; https://doi.org/10.1101/2025.07.08.663620doi: bioRxiv preprint Key laboratory of Carcinogenesis and Translational Research of Ministry of Education of China, 1031 Peking University Cancer Hospital & Institute, Beijing, China 1032 Aibin He 1033 Peking University Chengdu Academy for Advanced Interdisciplinary Biotechnologies, Chengdu, 1034 China 1035 Aibin He 1036 Contributions 1037 Y .S. and Y .W. conceived and designed this study. Y .W. and J.N. performed most biochemistry and 1038 cell biology experiments. W.Z. performed sequencing library construction and bioinformatics analysis. 1039 C. L. performed nucleosomes array reconstitution. X.W. performed super-resolution imaging 1040 quantification and analysis. P.X. performed cryo-electron tomography and data processing. W.Y . 1041 performed AFM. S. A. provided assistance with CUT&Tag experiments. Y .W., W.Z., J.N. and Y .S. 1042 wrote this paper. All authors participated in the discussion of the manuscript, and the manuscript was 1043 written through contributions of all authors. All authors have given approval to the final version of the 1044 manuscript. 1045 Corresponding authors 1046 Correspondence to Peng Xu, Cheng Li or Y ujie Sun. Further information and requests for resources 1047 and reagents should be directed to the lead contact, Y ujie Sun ([email protected]). 1048 Ethics declarations 1049 Competing interests 1050 The authors declare no competing interests. 1051 1052 was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted July 11, 2025. ; https://doi.org/10.1101/2025.07.08.663620doi: bioRxiv preprint Figures 1053 1054 Figure 1. NuMA regulates global 3D genome organization 1055 A. Schematic of auxin-inducible degron system (upper) and NuMA-depletion induced by auxin in 1056 unmodified HCT116 cells as control and HCT116-mAID-NuMA cells detected by western blotting 1057 (lower). 1058 B. IF imaging of Lamin A in untreated HCT116-mAID-NuMA cells as control and NuMA-depleted 1059 HCT116-mAID-NuMA cells induced by auxin. Scale bar, 10 μ m. 1060 C. FISH imaging (upper) and quantification (lower) of the volumes occupied by chromosome 2 1061 (green) and 18 (red) relative to the nucleus volume in untreated HCT116-mAID-NuMA cells as 1062 control and NuMA-depleted HCT116-mAID-NuMA cells induced by auxin. Error bars represent 1063 SD (n≥ 50). ****p < 0.0005, Mann-Whitney test. Scale bar, 5 μ m. 1064 D. Global chromatin accessibility upon NuMA-depletion detected by DNaseI digestion assay, with 1065 untreated HCT116-mAID-NuMA cells as control. Agarose gel image of genomic DNA digested by 1066 DNaseI at different concentrations (left), and percentages of high-molecular-weight (MW) 1067 genomic DNA (>5 kb) (right). The gel is representative of three independent experiments. Error 1068 bars represent SD (n=3). 1069 E. Heat map acquired by subtracting the inter-chromosomal ratios in untreated 1070 HCT116-mAID-NuMA cells as control and NuMA-depleted HCT116-mAID-NuMA cells induced 1071 by auxin (left). Trans-interaction ratios of each chromosome in control and NuMA-depletion 1072 induced by auxin in HCT116-mAID-NuMA cells (right). ***P < 0.001, paired t-test. 1073 F. Scatterplot showing compartment switches in untreated HCT116-mAID-NuMA cells as control 1074 and NuMA-depleted HCT116-mAID-NuMA cells induced by auxin. PC1 was calculated for each 1075 150 kb genomic segment to define A and B compartments and identify compartment switching, 1076 decompaction and compaction upon NuMA-depletion. 1077 G. A representative region showing the compartment switch from B to A upon NuMA-depletion. 1078 H. Heat map (resolution: 150 kb) acquired by subtracting compartment contacts in untreated 1079 HCT116-mAID-NuMA cells as control and NuMA-depleted HCT116-mAID-NuMA cells induced 1080 by auxin, and differential matrices of NuMA-depleted minus control cells. Below the heatmaps are 1081 PC1 values and gene density plots. 1082 I. Compartment A and B interaction ratios in untreated HCT116-mAID-NuMA cells as control and 1083 NuMA-depleted HCT116-mAID-NuMA cells induced by auxin. *P < 0.05, wilcoxon test for 1084 paired samples. 1085 1086 Figure 2. NuMA regulates heterochromatin compaction by maintaining 1087 nucleosome stacking 1088 A. A 15-state chromatin model established in HCT116 cells using ChromHMM. The color of each 1089 square represents the enrichment degree of chromatin feature. 1090 B. A TAC-seq fragment length and density of untreated HCT116-mAID-NuMA cells as control and 1091 NuMA-depleted HCT116-mAID-NuMA cells induced by auxin classified by chromatin types 1092 (Type 1, 2, and 3). Changes in nucleosome repeat length (NRL) and chromatin accessibility are 1093 was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted July 11, 2025. ; https://doi.org/10.1101/2025.07.08.663620doi: bioRxiv preprint shown for each chromatin state. The change in accessibility and NRL are shown as determined by 1094 NRLfinder. 1095 C. Chromatin accessibility upon NuMA-depletion in HCT116-mAID-NuMA cells as detected by 1096 Mnase digestion assay. Gel image of genomic DNA digested by Mnase at different concentrations 1097 (left), percentages of high MW genomic DNA (>5 kb) (middle), and an example of nucleosome 1098 ladders at the well 2 is shown (right). The gel is representative of three independent experiments. 1099 Error bars represent SD (n=3). 1100 D. STED imaging of H2B (left) and quantification (right) of nucleosome clutches in 1101 HCT116-mAID-NuMA cells, including the average area, frequency of nearest neighbor distance 1102 (NND) and average NND. Error bars represent SD (n=20). ****p < 0.0005, **p < 0.05, 1103 Mann-Whitney test. Scale bar, 5 μ m (left) and 1 μ m (right). 1104 E. A cartoon model depicting the changes in NRL and chromatin accessibility upon 1105 NuMA-depletion. 1106 1107 Figure 3. NuMA promotes linker histone H1’s binding to chromatin 1108 A． Representative live-cell image of over-expressed NuMA truncations and H1.1 in HeLa cells (left). 1109 Plots of the red and green pixel intensities along the white arrow in the left panel (right). Scale bar, 1110 5 μ m. 1111 B． Immunoblots showing the immunoprecipitation of HA-tagged NuMA and FLAG-tagged H1.1 1112 over-expressed in 293T cells. 1113 C． Heatmap profiling of H1’s CUT&Tag signal at NuMA-C’s regions, and the peaks were aligned 1114 using the center of the peaks. 1115 D． Representative examples of co-localization of NuMA-C and H1’s peaks. 1116 E． Heatmap profiling of H1’s CUT&Tag signal at NuMA-C’s regions in each chromatin type, and the 1117 peaks were aligned (left) using the center of the peaks and summed (right). 1118 F． CUT&Tag peaks of H1 in the whole genome and each chromatin type in untreated and 1119 NuMA-depleted HCT116-mAID-NuMA cells, and the peaks were aligned using the center of the 1120 peaks. 1121 G． Representative examples of H1 binding changes in each chromatin type. 1122 H． Representative live-cell image of over-expressed NuMA-C’s truncations and H1.1 in HeLa cells 1123 (left). Plots of the red and green pixel intensities along the white arrow in the left panel (right). 1124 Scale bar, 5 μ m. 1125 I． NuMA-C and H1.1 bind DNA together in vitro shown by EMSA analysis. 1126 was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted July 11, 2025. ; https://doi.org/10.1101/2025.07.08.663620doi: bioRxiv preprint Figure 4. NuMA facilitates nucleosome stacking by stabilizing H1 1127 A. The binding preference of NuMA-C to nucleosomes with H1 as detected by mono-nucleosome 1128 pull-down experiment and western blotting. 1129 B. AFM images for the reconstituted 12× nucleosome arrays with H1.4 and NuMA-C. Scale bar, 100 1130 nm. 1131 C. Sucrose density gradient centrifugation result of 12 × nucleosome array with H1.4 and NuMA-C. 1132 DNA displayed by EB staining. 1133 D. Upper left, 3D rendering from tomograms of U2OS cells overexpressing full-length NuMA 1134 (NuMA-FL) Lower left, 3D rendering from tomograms truncation deleting the C-terminal domain 1135 (NuMA-dC). Scale bar, 100 nm. Upper right, in situ nucleosome subtomogram averaging electron 1136 density map from cells overexpressing NuMA-FL. Lower right, in situ nucleosome subtomogram 1137 averaging electron density map from cells overexpressing NuMA-dC. H1 linker region labeled 1138 orange. Features labeled with black arrows. 1139 E. In situ nucleosome NND distribution. **p < 0.05, Mann-Whitney test. 1140 1141 Figure 5. NuMA oligomerizes into quasi-network organization through its 1142 C-termini in vivo 1143 A. Schematic diagram of antibody epitopes targeting NuMA-N and NuMA-C. 1144 B. Expanded 2D IF images of NuMA-N and NuMA-C and reconstructed NuMA oligomers in U2OS 1145 cells. NuMA-N was labeled by Atto647N (magenta) and NuMA-C was labeled by Alexa594 1146 (cyan). Scale bars, 5 μ m and 250 nm. 1147 C. Calculated ratio of numbers (upper) and distribution of nearest neighbor distance (NND) from 1148 NuMA-C to NuMA-N clusters in U2OS 2D expanded IF images (lower). 1149 D. Perspective view (left) and orthogonal view (right) of expanded 3D IF images of NuMA-N and 1150 NuMA-C in U2OS cells. Scale bars, 5 μ m and 2 μ m. 1151 E. Perspective 3D view and crop of reconstructed NuMA oligomers in U2OS 3D expanded IF images. 1152 Scale bars, 5 μ m and 1 μ m. 1153 F. Calculated ratio of numbers (left) and distribution of NND from NuMA-C to NuMA-N clusters in 1154 U2OS 3D expanded IF images (right). 1155 G. Expanded IF images of H3K9me3, H3K27me3 and H3K4me3 co-stained with NuMA-C in U2OS 1156 cells. Scale bars, 5 μ m and 1 μ m. Error bars represent SD (n=10). ****p < 0.0005, Mann-Whitney 1157 test. 1158 H. Expanded IF images of H1 co-stained with NuMA-C in U2OS cells (left) and Pearson’s coefficient 1159 of NuMA-C and H1 (right). The NuMA-C clusters were randomized and calculated as control. 1160 Scale bar, 5 μ m. 1161 1162 Figure 6. NuMA contributes to epigenetic maintenance of constitutive 1163 heterochromatin and repression of LTR expression 1164 A. IF imaging (left) and quantification of the heterogenous index (right) of H3K9me3, H3K27me3 1165 and H3K4me3 in HCT116-mAID-NuMA cells after induced by auxin, with untreated 1166 was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted July 11, 2025. ; https://doi.org/10.1101/2025.07.08.663620doi: bioRxiv preprint HCT116-mAID-NuMA cells as control. Error bars represent SD (n ≥ 50). ****p < 0.0005, 1167 Mann-Whitney test. Scale bar, 5 μ m. 1168 B. Representative expanded IF images of histone H1, NuMA co-stained with H3K9me3, H3K27me3 1169 and H3K4me3 (left) in NuMA-depleted HCT116-mAID-NuMA cells and untreated 1170 HCT116-mAID-NuMA cells as control. Pearson’s coefficient of histone H1 with H3K9me3, 1171 H3K27me3 and H3K4me3 (right). Scale bar, 5 μ m. 1172 C. Fraction of H3K9me3 and H3K27me3 enrichment in genomic regions where NuMA-C and H1 1173 bind and overlap, and among the whole genome. 1174 D. Percentage of genomic regions with up-, down- and unchanged peaks of H3K27me3 (upper) and 1175 H3K9me3 (lower) upon NuMA-depletion. 1176 E. Fraction of NuMA-C and H1 enrichment in genomic regions with unchanged H3K9me3, 1177 down-regulated H3K9me3, up-regulated H3K9me3 upon NuMA-depletion and in the whole 1178 genome. The peaks were aligned using the center of the peaks. 1179 F. Changes of H3K9me3 CUT&Tag peaks upon NuMA-depletion in different categories (regions 1180 where NuMA-C and H1 are co-bound and regions bound by NuMA or H1 alone), with the change 1181 profile in the whole genome as control. 1182 G. Percentage of multiple types of transposable elements in all genomic regions with unchanged 1183 H3K9me3 and down-regulated H3K9me3 upon NuMA-depletion. 1184 H. Expression changes of LTRs in NuMA-C and H1 co-bound regions with unchanged and 1185 down-regulated H3K9me3 upon NuMA-depletion. 1186 I. Representative example of expression up-regulation in genomic regions where NuMA-C and H1 1187 are co-bound and with down-regulated H3K9me3 upon NuMA-depletion. 1188 1189 Figure 7. Models for NuMA’s promotion on constitutive heterochromatin 1190 compaction through stabilizing linker histone H1 1191 NuMA interacts with H1 with its C-terminus, stabilizes H1’s binding to DNA, enhance H1’s chromatin 1192 compaction effect and nucleosome stacking, then contributes to epigenetic maintenance of constitutive 1193 heterochromatin and repression of LTR expression. 1194 was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted July 11, 2025. ; https://doi.org/10.1101/2025.07.08.663620doi: bioRxiv preprint was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted July 11, 2025. ; https://doi.org/10.1101/2025.07.08.663620doi: bioRxiv preprint was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted July 11, 2025. ; https://doi.org/10.1101/2025.07.08.663620doi: bioRxiv preprint was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted July 11, 2025. ; https://doi.org/10.1101/2025.07.08.663620doi: bioRxiv preprint was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted July 11, 2025. ; https://doi.org/10.1101/2025.07.08.663620doi: bioRxiv preprint was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted July 11, 2025. ; https://doi.org/10.1101/2025.07.08.663620doi: bioRxiv preprint was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted July 11, 2025. ; https://doi.org/10.1101/2025.07.08.663620doi: bioRxiv preprint was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted July 11, 2025. ; https://doi.org/10.1101/2025.07.08.663620doi: bioRxiv preprint