Two-step regulation of centromere distribution by condensin II and the nuclear envelope proteins

The arrangement of centromeres within the nucleus differs among species and cell types. However, neither the mechanisms determining centromere distribution nor its biological significance are currently well understood. In this study, we demonstrate the importance of centromere distribution for the maintenance of genome integrity through the cytogenic and molecular analysis of mutants defective in centromere distribution. We propose a two-step regulatory mechanism that shapes the non-Rabl-like centromere distribution in Arabidopsis thaliana through condensin II and the linker of the nucleoskeleton and cytoskeleton (LINC) complex. Condensin II is enriched at centromeres and, in cooperation with the LINC complex, induces the scattering of centromeres around the nuclear periphery during late anaphase/telophase. After entering interphase, the positions of the scattered centromeres are then stabilized by nuclear lamina proteins of the CROWDED NUCLEI (CRWN) family. We also found that, despite their strong impact on centromere distribution, condensin II and CRWN proteins have little effect on chromatin organization involved in the control of gene expression, indicating a robustness of chromatin organization regardless of the type of centromere distribution. In Arabidopsis cells undergoing mitosis, centromere distribution is shown to be regulated by two steps: scattering in M-phase and stabilization in interphase. This may affect the maintenance of genome integrity rather than gene regulation.

T he centromere is a specialized chromosomal domain that is universally conserved in all eukaryotic chromosomes and is indispensable for the equipartition of chromosomes during mitosis. The spatial arrangement of centromeres within the nuclear space in somatic interphase differs among species, cell types, cell-cycle stages and differentiation states [1][2][3] . Centromere distribution is classified into two types: Rabl and non-Rabl configurations [1][2][3][4] . In the Rabl configuration, the centromeres are grouped at one side of the nucleus, whereas the telomeres, constituting a universal constitutive heterochromatic domain, are clustered in the opposite nuclear hemisphere. By contrast, in the non-Rabl configuration, the centromeres and telomeres are dispersed uniformly within the nucleus 5 . The Rabl configuration can be considered the 'default' organization of interphase chromatin because it retains the orientation of chromosomes that segregate during anaphase 3 . Therefore, mechanisms acting after anaphase are required to establish a non-Rabl centromere configuration; 6 however, the underlying processes are yet to be fully understood, particularly in plants.
It has been reported that the activity of condensin II, composed of two structural maintenance of chromosome (SMC) proteins and three non-SMC proteins (CAP-D3, CAP-G2 and CAP-H2), prevents centromere clustering in human and Drosophila melanogaster cells 7,8 . In these cells, condensin II-mediated lengthwise chromosome compaction during mitosis or interphase has been proposed as a model to prevent centromere clustering and, consequently, establish a non-Rabl centromere configuration 7,8 . In Arabidopsis thaliana, the centromeres are scattered in the nucleus near the nuclear periphery, adopting a non-Rabl configuration 9,10 Similar to animal cells, condensin II is required to prevent the hyperclustering of centromeres in A. thaliana [11][12][13] . However, it remains unclear whether this is based on the same molecular mechanism(s).
In addition to condensin II, several other factors involved in the spatial arrangement of centromeres have been reported in A. thaliana. Among the CROWDED NUCLEI (CRWN) proteins, which form a meshwork structure at the nuclear lamina 14 , CRWN1, CRWN2 and CRWN4 function in the unclustering of chromocentres. The latter are densely compacted heterochromatic regions containing centromeric and pericentromeric DNA [14][15][16][17] . In addition, CRWN1 interacts directly with heterochromatin, including centromeric and pericentromeric regions 18 . The Sad1/UNC-84 (SUN) domain proteins, subunits of the linker of the nucleoskeleton and cytoskeleton (LINC) complex in the nuclear membrane, play a role in maintaining centromeres near the nuclear periphery 16 . However, the role of these factors in establishing the interphase centromere distribution is not well understood. Although these factors, including condensin II, have similar roles in the spatial arrangement of centromeres, it remains unclear whether they function in a coordinated and concerted manner or independently of each other.
In this study, we demonstrate the concerted action of condensin II and the LINC complex to determine centromere distribution during mitosis. In addition, we found that CRWN1 and CRWN4 are required to stabilize centromere distribution during interphase. The arrangement of centromeres within the nucleus differs among species and cell types. However, neither the mechanisms determining centromere distribution nor its biological significance are currently well understood. In this study, we demonstrate the importance of centromere distribution for the maintenance of genome integrity through the cytogenic and molecular analysis of mutants defective in centromere distribution. We propose a two-step regulatory mechanism that shapes the non-Rabl-like centromere distribution in Arabidopsis thaliana through condensin II and the linker of the nucleoskeleton and cytoskeleton (LINC) complex. Condensin II is enriched at centromeres and, in cooperation with the LINC complex, induces the scattering of centromeres around the nuclear periphery during late anaphase/telophase. After entering interphase, the positions of the scattered centromeres are then stabilized by nuclear lamina proteins of the CROWDED NUCLEI (CRWN) family. We also found that, despite their strong impact on centromere distribution, condensin II and CRWN proteins have little effect on chromatin organization involved in the control of gene expression, indicating a robustness of chromatin organization regardless of the type of centromere distribution.

Two-step regulation of centromere distribution by condensin II and the nuclear envelope proteins
Our work reveals a coordinated mechanism consisting of two steps that enable a non-Rabl-type centromere distribution. Moreover, the characterization of mutants affecting centromere distribution implies a possible role of centromere configuration in maintaining genome integrity, but not chromatin organization, which is associated with the regulation of gene expression.

Phase-specific enrichment of condensin II at centromeres
To further characterize the function of condensin II, its subnuclear localization was visualized using green fluorescent protein (GFP) fused to CAP-G2, a subunit specific to condensin II, in wild type A. thaliana nuclei. Condensin II is strongly expressed in root tips; 19 hence, we decided to focus our investigation on cells in the meristematic, elongation and differentiation zones (DZs) of the root. In all regions, CAP-G2-GFP signals were detected in both the nucleoplasm and nucleolus (Fig. 1a). Moreover, we found that some nuclei showed a spot-like localization of condensin II only in the meristematic zone (MZ), in addition to its basic localization throughout the nucleus (Fig. 1a). Other GFP-fused condensin II-specific subunits, CAP-H2 and D3, did not show stable nucleolar expression, but nucleoplasmic and spot-like localizations were common ( Supplementary Fig. 1a). Given that the cells in the MZ proliferate, it is possible that the formation of a spot-like localization of condensin II is cell-cycle dependent. Indeed, using time-lapse imaging analysis of CAP-G2-GFP during the mitotic cell cycle, the spot-like localization of condensin II was observed 80 min before metaphase and disappeared around 40 min after metaphase (Fig. 1b). According to previous live imaging studies of cell-cycle progression in the root meristem 20,21 , these time points correspond to the late G2 and early G1 phases, respectively.
In the primitive red alga Cyanidioschyzon merolae and the nematode Caenorhabditis elegans, condensin II is specifically localized to centromeres from prophase to anaphase 22,23 . Thus, we hypothesized that the condensin II spots in A. thaliana nuclei co-localize with centromeres. Ideally, mitotic cells display 10 discrete centromere signals, which can be visualized using the fluorescent fusion protein tdTomato-CENH3 (centromeric histone H3). As expected, our observations revealed that the spots of CAP-G2-GFP, CAP-H2-GFP and CAP-D3-GFP co-localized with the tdTomato-CENH3 signal ( Fig. 1c and Supplementary Fig. 1b). Collectively, these results suggest that condensin II accumulates at centromeres during this specific cell-cycle phase.

Condensin II establishes a scattered centromere distribution
Fluorescent in situ hybridization (FISH) analysis has shown that condensin II is indispensable for the distribution of centromeres in interphase nuclei 11,12 . To determine how condensin II participates in the dynamic regulation of centromeres during the cell cycle, we conducted live imaging of tdTomato-CENH3 in the cap-h2-2 (ref. 19 ) null mutant. Contrary to the scattered centromere distribution in the wild type, the cap-h2-2 mutant showed a polarized centromere distribution, in which almost all centromeres were clustered at one side of the nucleus (Fig. 2a). This occurred not only in meristematic cells, but also in differentiated cells of the root, guard cells and leaf epidermal cells ( Fig. 2a and Supplementary Fig. 1c). Next, we quantified the polarized centromere distribution in the root meristem. Given that cells above the stem cell niche divide longitudinally into the root apical meristem, we determined an equatorial plane that produces two hemispheres in the nucleus, using the centre points of two nuclei neighbouring the target nucleus in the same cell file. We calculated the degree of polarization in centromere distribution by counting the number of centromeres in each hemisphere (Fig. 2b, left). The cap-h2-2 mutant showed a clearly polarized centromere distribution compared with the wild type (Fig. 2b,  right). We also confirmed a similar centromere distribution in the mutants of other condensin II-specific components, cap-g2-1 and cap-d3-1 ( Supplementary Fig. 1d,e). Furthermore, we showed that the phenotypes of all condensin II mutants were complemented by expression of the corresponding components fused to GFP from a transgene ( Supplementary Fig. 1b,d). These results establish a requirement of the condensin II 'complex' for the scattering of centromeres.
To determine the period during which centromere positioning mediated by condensin II was established, we monitored centromere dynamics from anaphase to early G1 phase (Fig. 2c). In the wild type, all centromere pairs were aligned along the metaphase plate. Sister centromeres then moved to opposite poles in anaphase, subsequently scattering in telophase ( Fig. 2c and Supplementary Fig. 2) 9,24 . Our imaging analysis revealed that, during the transition from anaphase to telophase, the centromeres remained at opposite poles in the cap-h2-2 mutant. This position persisted after entry into the G1 phase ( Fig. 2c and Supplementary  Fig. 2). Given that condensin II was localized to centromeres throughout mitosis (Fig. 1), condensin II function at centromeres during the transition from anaphase to telophase seems indispensable for centromere scattering.
Next, we evaluated whether condensin II plays a role in establishing centromere peripheral localization, by visualizing the centromere and nuclear periphery using tdTomato-CENH3 and SUN1 fused to the yellow fluorescent protein (SUN1-YFP), respectively. We found that even in cap-h2-2, polarized centromeres showed peripheral localization, which is consistent with the results of a previous study on cap-d3 mutants 13 .
Intriguingly, the distribution pattern of centromeres in condensin II mutants resembles the Rabl centromere configuration. This led us to speculate that telomere positioning was also affected in these mutants. However, our previous FISH analysis showed that defects in condensin II did not affect telomere localization to the nucleolar periphery in A. thaliana 11 . This result was confirmed by simultaneous visualization of telomeres and centromeres using a telomere-binding protein, conserved telomere maintenance component 1 (CTC1)-tdTomato 25 and VENUS-CENH3, respectively ( Supplementary Fig. 3b).
Previously, we have shown that, in condensin II mutants, there were nuclei with 45S ribosomal DNA signal dissociation at three or four centromeres 11 . Two of the four 45S rDNA signals are on chromosome 4 and localized to the nucleolus 26 , suggesting that the distribution of centromeres localized to the nucleolus is also affected in the condensin II mutants. This raised the question of whether the position of the nucleolus in the nucleus is affected in the condensin II mutant. Therefore, we analysed the position of the nucleolus; however, we found no difference between the wild type and condensin II mutants ( Supplementary Fig. 3c-e).
Taken together, these results suggest that condensin II is not involved in the peripheral localization of centromeres or the positioning of nucleoli and telomeres, and that centromere scattering is regulated independently of their positioning.

Centromeres are scattered by the condensin II-lINC complex
Nuclear envelope (NE) reassembly occurs in plants during the transition from anaphase to telophase of mitosis 24,27 . In A. thaliana, at late anaphase, the NE proteins SUN1 and SUN2 are reassembled at the distal surfaces of the chromosomes, where centromeres are found. These proteins enclose the chromosomes from the distal surface to the proximal surface at telophase 24 . Similar dynamics have been observed for plant nuclear lamina proteins, namely nuclear matrix constituent proteins, in Apium graveolens (celery) 27 . Because Next, the transverse axis (orange dotted line) was drawn perpendicular to the longitudinal axis through the centroid of the target nucleus, dividing it into two equal hemispheres. The number of centromeres (magenta circles) in each hemisphere was then counted and divided by the total number of separated centromeres in each nucleus. Between the two values obtained from the two hemispheres in each nucleus, a higher value was adopted as the polarization level. As the lowest theoretical value is 0.5, we subtracted from 0.5 the obtained value for display in the graph, which shows the mean of the values for the polarization level ± s.e.m., n = 23 nuclei from at least three independent plants, two-sided Student's t-test. c, Dynamics of centromeres visualized by p35S::CENH3-tdTomato from metaphase to early G1 phase. Nuclei are visualized in green by pRPS5a::H2B-GFP. The optical section closest to the centre of the nucleus is shown. Scale bars, 5 μm.
the timing between the reassembly of NE-related factors and centromere scattering is approximately the same, we hypothesized that NE-related factors are involved in centromere scattering. The SUN proteins (SUN1-SUN5) are inner nuclear membrane proteins comprising the LINC complex in association with other proteins 28 . SUNs have the potential to interact with chromatin in the nucleoplasm, along with their interactions with Klarsicht/ ANC-1/Syne homology (KASH)-domain proteins, such as WPP domain-interacting proteins and SUN-interacting NE proteins in the outer nuclear membrane. Through interactions between KASH or KASH-associated proteins, such as WPP domain-interacting tail-anchored proteins and myosin XI-I, with actin filaments, the SUNs are also connected to the cytoskeleton 28 . The nuclear lamina comprising A. thaliana nuclear matrix constituent proteins, known as CRWN proteins, is located underneath the NE (Fig. 3a) 28,29 . To evaluate the involvement of NE-related factors in centromere scattering, we analysed centromere distribution in mutants affecting the LINC complex and CRWN proteins (Fig. 3b-e). Mutants affecting KASH (wifi and sine1-1) and SUN proteins (sun1-KO sun2-KD and sun4 sun5) showed a significantly more polarized centromere distribution than the wild type, although the polarization was not as pronounced as observed previously in the condensin II mutants. By contrast, centromere distribution in the crwn1 crwn4 (hereafter crwn1/4) and myosin xi-i mutants was comparable to that in the wild type (Fig. 3b,c). In addition, we found that the inhibition of actin polymerization by latrunculin B (Lat B) treatment caused a polarized centromere distribution, whereas treatment with oryzalin (Ory), resulting in microtubule depolymerization, did not (Fig. 3d,e). These results suggest that, in addition to condensin II, the LINC complex associated with actin filaments is involved in determining centromere distribution during mitosis.
Next, we investigated whether the LINC complex associates with the centromere and/or condensin II to regulate centromere distribution. Among the LINC complex components, only SUN proteins have the potential to interact with chromatin as part of the N-terminus in the nucleoplasm 28 . Therefore, we evaluated the interaction of SUN proteins with CENH3 and the condensin II components CAP-H2 and CAP-G2, using co-immunoprecipitation (Co-IP). After transient expression of the proteins in tobacco leaves, our Co-IP analyses indicated interactions of SUN1 and SUN2 with CENH3 ( Fig. 3f). Similarly, we found an interaction between SUN2 and the condensin II component CAP-G2 (Fig. 3g). The interactions between the SUNs and condensin II components were further supported by yeast two-hybrid assays (Fig. 3h). Collectively, these results suggest that the LINC complex is essential for correct centromere distribution in interphase through interactions between SUNs, centromeres and condensin II, most probably during the transition from anaphase to telophase. The corresponding mutants did not show a marked alteration in centromere distribution, as was observed in cap-h2-2, suggesting functional redundancy among the SUN and KASH proteins, respectively. Hereafter, we termed the interaction of condensin II with the LINC complex involved in the regulation of centromere positioning, CII-LINC.

CrWNs stabilize the position of scattered centromeres
As mentioned earlier, no polarization in centromere distribution was observed in crwn1/4 mutants by imaging (Fig. 3b,c). However, when the number of distinguishable centromeres in interphase nuclei was counted, the crwn1/4 nuclei had a significantly lower number of centromeres than the wild type, similar to the cap-h2-2 condensin II mutant ( Supplementary Fig. 4) and consistent with a previous study using FISH 11,14 . These findings imply that CRWNs contribute to centromere positioning in a manner different from condensin II. To date, CRWNs have been shown to function in tethering chromatin, including pericentromeric and centromeric heterochromatin, to the nuclear periphery 18 . Consistent with this, we found CRWN1 to interact with CENH3 in Co-IP assays in tobacco leaves and an enrichment of CRWN1-GFP at tdTomato-CENH3 positive foci in interphase nuclei (Fig. 4a,b). These findings support the idea that CRWNs directly regulate centromere positioning.
To obtain further insights into the function of CRWNs in centromere positioning, we analysed centromere movement in interphase nuclei. Live imaging demonstrated that centromeres are highly dynamic in crwn1/4 mutant nuclei, but static in wild type and mutant plants affecting CII-LINC; that is, cap-h2-2 and sun1-KO sun2-KD, which show a polarized centromere distribution (Fig. 4c,d). This result indicates that CRWNs restrain the positions of centromeres during interphase at the nuclear periphery. Next, to evaluate the genetic interaction between CII-LINC and CRWNs in the regulation of centromere positioning, we observed centromere dynamics in the triple cap-h2-2 crwn1/4 mutant. Intriguingly, we found both types of centromere distribution, polarized and scattered, in this triple mutant (Fig. 5a). Subsequently, we monitored centromere dynamics in the triple mutant (Fig. 5b). First, the triple mutant clearly showed a polarized centromere distribution, as observed in cap-h2-2 ( Fig. 2c, from 15 to 60 min in Fig. 5b). Over time, however, the centromeres started to move around (after 80 min) and, consequently, adopted a distribution similar to that observed in the wild type (from 110 to 125 min in Fig. 5b), in contrast to the polarized distribution in cap-h2-2 (Figs. 2c and 4c). Consistently, a quantitative analysis of centromere distribution revealed a gradual shift from polarized to scattered centromeres (Fig. 5c). Even after centromere scattering, centromere movement in the triple mutant was more dynamic than that in wild type or cap-h2-2 nuclei (Fig. 4c,d). These observations suggest that the NE-stabilization of centromeres by CRWNs occurs continuously during interphase, and that CII-LINC and CRWNs function independently of each other to regulate centromere positioning.
Considering this, we propose that centromere distribution is determined by a two-step process after the centromere pairs segregate at early anaphase: (i) interactions of CII-LINC with centromeres occur at late anaphase, mediating the scattering of centromeres from late anaphase to telophase; and (ii) CRWNs stabilize the position of scattered centromeres after the entry into interphase (Fig. 5d).

roles of condensin II and CrWNs in chromatin organization
Considering that both CII-LINC and CRWNs have a major impact on centromere distribution, we investigated whether deficiencies in these factors alter higher-order chromatin organization. We performed duplicate Hi-C analysis in wild type, cap-h2-2, crwn1/4 and cap-h2-2 crwn1/4 plants (Supplementary Table 2). A high correlation between the duplicates was obtained ( Supplementary Fig. 5a,b), and the combined data from duplicate experiments were used for subsequent analyses to improve resolution. The map of Hi-C contact frequencies in cap-h2-2 did not exhibit an X-shaped trans-interaction signal ( Supplementary Fig. 6), a feature of chromatin interactions observed in the Hi-C contact map of barley, in which chromosomes display a typical Rabl configuration 30 . To compare contact frequencies, we calculated the relative differences between all elements of the two Hi-C matrices of interest, as described previously 31 (Fig. 6a). Using visual inspection, we found that cap-h2-2 exhibited increased inter-chromosomal pericentromere contacts and increased contacts between the two centromere-flanking halves of the pericentromeres. We also observed elevated cis-chromosomal inter-arm and trans-chromosomal inter-arm contacts. The cap-h2-2 mutant also showed a conspicuous decrease in pericentromerearm contacts (both cis-and trans-chromosomal), and a slight decrease in intra-arm contacts and interactions within centromeres ( Supplementary Fig. 7). Similar alterations in chromatin contacts were observed in the crwn1/4 double and cap-h2-2 crwn1/4 triple mutants; however, their overall magnitude was greater than that observed in cap-h2-2 ( Supplementary Fig. 8). These results indicate that the defects in both condensin II and CRWN mutants have similar effects on chromatin organization, albeit to varying degrees. In addition, a comparison of crwn1/4 and cap-h2-2 crwn1/4 showed only slightly more enhanced differences in some interacting regions, including inter-chromosomal pericentromere interactions and interactions within each centromere, which are probably caused by the additional mutation in CAP-H2 ( Supplementary Fig. 8). These findings suggest that, although condensin II and CRWNs have similar functions in organizing chromatin interactions, at least in specific regions, they work independently of each other.
Next, we analysed the formation of discrete structural domains by calculating the correlation coefficients of the distance-normalized interaction matrix 31 (Fig. 6b). The correlation matrix derived from a Hi-C map is closely related to how strong interactions/ depletions are among chromatin regions and helps to highlight the structural separation between different chromosomal regions. Therefore, a weakened correlation matrix indicates a lower degree of spatial separation of the two chromatin compartments, A and B 18 .  The cap-h2-2 Hi-C map showed a subtly weaker correlation matrix than the wild type, indicating slightly less well-defined chromatin compartmentalization. A further weakened chromatin compartmentalization was observed in the crwn1/4 Hi-C map, as reported previously for the individual mutants, crwn1 and crwn4 (ref. 18 ), and in the cap-h2-2 crwn1/4 Hi-C map (Fig. 6b). We also analysed the number of contacts within and between compartments (Fig. 6c) and found reduced contacts within and increased contacts between compartments, respectively, supporting the idea of weakened chromatin compartmentalization in the mutants. Next, we performed a principal component analysis on the correlation matrix of each chromosome, including centromeric regions. In this analysis, regions exhibiting negative eigenvalues corresponded to B compartments that, in A. thaliana, mainly cover constitutive heterochromatin of pericentromeres. We found that crwn1/4 and cap-h2-2 crwn1/ 4, but not cap-h2-2, showed a high incidence of compartment switches from A to B (Fig. 6d). In addition, both crwn1/4 and cap-h2-2 crwn1/4 exhibited a similarly altered eigenvalue distribution pattern restricted to the pericentromeric regions of all chromosomes, with negative eigenvalues (B compartment) expanding into more distal regions of the pericentromeres. Consequently, the eigenvalue boundaries between pericentromeres and chromosome arms were shifted distally towards the chromosome arms in those mutants (Fig. 6e). To further evaluate chromatin structure, we determined the interaction decay exponents (IDEs) of entire chromosomes, pericentromeres and chromosome arms (Fig. 6f). The IDE is indicative of the chromatin structure model, that is, whether a region behaves according to the fractal-or the equilibrium-globule model 32 . No differences in the IDEs of the entire chromosomes were observed between the mutants and the wild type. However, significantly higher pericentromeric IDEs were observed in crwn1/4 and cap-h2-2 crwn1/4 mutants, with IDEs close to those of chromosome arms (Fig. 6f). This suggests a 'decompaction' of parts of the pericentromeres in these mutants. On the other hand, no major alterations were observed in these mutants with respect to the IDEs of chromosome arms (Fig. 6f). Next, the distance-to-local contact ratio (DLR) was calculated to evaluate the compactness of chromatin on chromosomes. The DLR represents the ratio of local to distant chromatin contact frequencies in a given chromosomal region and can be used as a proxy for chromatin compactness (Fig. 6g). Comparison of the DLRs in chromosome 5 between the wild type and crwn1/4 or cap-h2-2 crwn1/4, respectively, showed less compaction of pericentromeric chromatin in those mutants, consistent with the results on the IDEs. Together with the reduced interaction between the pericentromere and the chromosome arms (Fig. 6a,g), this result suggests that the centromere core region is spatially distant from the chromosome arms in these mutants. Interestingly, this decompaction of pericentromeric chromatin is not accompanied by substantial changes in DNA methylation at pericentromeres (Supplementary Fig. 9). Furthermore, it is worth noting that the eigenvalue distribution, the IDEs of pericentromeres and the DLRs were comparable between cap-h2-2 and the wild type (Fig. 6d-f). Collectively, these results suggest that CRWNs are involved in the organization of chromatin structure, especially at pericentromeres, establishing a defined compartmentalization of heterochromatin. By contrast, the contribution of condensin II to chromatin structure was not evident, despite its involvement in chromatin interactions at pericentromeres. Previous studies reported that the transcriptome is not substantially affected in a crwn1 single mutant, which shows weakened chromatin compartmentalization and higher IDEs at pericentromeres 18,31 similar to the crwn1/4 double mutant used in this study. Transcriptome analyses of the crwn1/4 double mutant detected only 131 (microarray) and 128 (RNA sequencing (RNA-seq)) differentially expressed genes (DEGs), which had no specificity in their chromosomal location (Supplementary Figs. 10 and 11). Similarly, the cap-h2-2 mutation had little effect on the transcriptome; there were only 131 (microarray) and 119 (RNA-seq) DEGs ( Supplementary  Figs. 10 and 11). These results suggest that, in addition to alterations in chromatin organization, abnormal centromere distribution does not strongly influence local gene regulation in mutants affecting CRWNs or condensin II. By contrast, 867 DEGs (RNA-seq) were found in the cap-h2-2 crwn1/4 triple mutant ( Supplementary  Fig. 11). Given that there were no major differences in the chromatin organization of crwn1/4 and cap-h2-2 crwn1/4, the increased number of significantly changed genes in cap-h2-2 crwn1/4 is probably due to other effects brought about by the genetic interaction between cap-h2-2 and crwn1/4.

Biological significance of regulating centromere distribution
To evaluate the biological significance of CII-LINC-and CRWNmediated centromere distribution, we focused on investigating its potential role in maintaining genome integrity. Condensin II alleviates DNA damage caused by abiotic stress and genotoxic chemical treatments 19 . CRWNs, including CRWN1 and CRWN4, are involved in protecting genomic DNA against oxidative stress caused by methyl methanesulfonate 33,34 . Accordingly, we evaluated the sensitivity of primary root growth in mutants showing abnormal centromere distribution to the induction of DNA double-strand breaks (DSBs) upon treatment with the DSB-inducing reagent zeocin. Defects in primary root growth and root morphology were found in the mutants cap-h2-2, crwn1/4 and sun1/2 (a double-knockout mutant in a Col-0/Ws heterozygous background) after 2 or 4 days of treatment with zeocin ( Fig. 7a and Supplementary Fig. 12). In addition, comet assays revealed that all mutants had increased DSB levels compared with the wild type under both normal and DSBinducing conditions (Fig. 7b). These results confirm that genome integrity is impaired in mutants with an abnormal centromere   distribution. In addition, we found that, compared with cap-h2-2 and crwn1/4, the cap-h2-2 crwn1/4 mutant displayed an additive phenotype with respect to root growth, morphology and DSB levels (Fig. 7a,b and Supplementary Fig. 12), suggesting that condensin II and CRWNs contribute to the maintenance of genome integrity via different mechanisms.
Next, we performed immunostaining of phosphorylated histone H2AX (γH2AX) foci, providing information on the extent and location of DNA damage based on the number and location of foci detected in the nucleus 33 (Fig. 7c). We confirmed that the number of γH2AX foci was significantly increased in the cap-h2-2, crwn1/4 and cap-h2-2 crwn1/4 mutants compared with the wild Statistical analyses were performed for each treatment group, n ≥ 120 nuclei, P < 0.05, one-way ANOVA and Tukey's HSD. Different letters indicate that the values are statistically different. c, Representative immunofluorescence images of γH2AX (magenta) in the nuclei of root tip cells from cap-h2-2 crwn1/4 treated with or without zeocin. The nuclei were stained with DAPI (cyan). The images are shown as maximum z-projections. Scale bars, 5 μm. d, Number of γH2AX foci in the nuclei of the root tips with and without zeocin treatment (2.5 μM). The frequency distribution of the number of γH2AX foci in each mutant was statistically compared to that in Col-0 using a two-sided Fisher's exact test (*P < 0.05, **P < 0.01, n ≥ 315 nuclei). e, Representative immunofluorescence image of γH2AX (magenta) co-localized with a chromocentre (cyan, arrowhead) in the nucleus of a root tip cell from cap-h2-2 crwn1/4 treated with zeocin. The optical section closest to the centre of the nucleus is shown. Scale bar, 5 μm. f, Proportion of nuclei with γH2AX foci detected in the chromocentres of root tip nuclei with and without zeocin treatment (2.5 μM). Probabilities for nuclei with γH2AX foci co-localizing with a chromocentre were compared between the wild type and mutants using a two-sided Fisher's exact test (*P < 0.01, n ≥ 315 nuclei). Precise values of n are shown as discrete numbers in a, b, d and f. type (Fig. 7d). Under our conditions, the frequency of γH2AX foci co-localizing with 4′,6-diamidino-2-phenylindole (DAPI)-stained chromosomes was significantly higher in all mutants (Fig. 7e,f), indicating that the integrity of centromeric and pericentromeric chromatin was compromised.
To assess whether the increased sensitivity of the mutants to DNA damage could be attributed to impaired gene regulation in response to the induction of DNA damage, we performed RNA-seq analysis 0, 3 and 48 h after the start of zeocin treatment in cap-h2-2, crwn1/4 and cap-h2-2 crwn1/4 mutants (Supplementary Fig. 11). After 3 h of zeocin treatment, the number of DSB-responsive DEGs was very low in cap-h2-2 (2 DEGs) and crwn1/4 (76 DEGs). The same trend was observed after 48 h (44 and 74 DEGs in cap-h2-2 and crwn1/4, respectively), when a clear difference in the extent of DNA damage was observed between the wild type and the mutants. By contrast, we found more DEGs in cap-h2-2 crwn1/4 at both 3 h (336 DEGs) and 48 h (250 DEGs). However, GO enrichment analysis revealed that no GO terms related to the response to DNA damage or DNA repair were enriched among these DEGs (Supplementary Data 3). Taken together, these results suggest that the transcriptional response to DNA damage is maintained almost normally in all mutants tested.

Discussion
In comparison with most plants in which centromere distribution adopts a Rabl configuration in interphase, A. thaliana centromeres were considered to be distributed randomly 35 . However, previous studies using a spatial statistics approach found a more regular distribution of centromeres than expected under the assumption of randomness. Interactions of centromeres with certain nuclear factors, including the NE and nucleolus, have been proposed to contribute to this specific distribution 26,36,37 . Here, we reveal a new mechanism for centromere distribution in interphase nuclei that can be divided into two steps: (i) a scattering step dependent on interactions of the centromere with the CII-LINC complex during mitosis; and (ii) a stabilizing step dependent on interactions of the centromere with the nuclear lamina throughout interphase (Fig. 5c). This two-step process contributes substantially to the spatial regularity of centromeres in A. thaliana.
We propose that condensin II mediates the association of centromeres with the NE during mitosis by interacting with the LINC complex, playing an essential role in centromere scattering in A. thaliana (Fig. 5c). A different model was proposed for human cells, in which the lengthwise compaction of chromosomes by condensin II during mitosis determines the interphase centromere distribution 8 . In addition, a role for interphase condensin II in regulating centromere distribution has been reported in D. melanogaster. Enhanced interphase condensin II activity drives the force that separates and scatters clustered centromeres around the nuclear periphery by promoting the lengthwise compaction of chromosomes in the polyploid nuclei of nurse cells 7 . Either way, condensin II has a determining role in regulating centromere distribution in eukaryotes including A. thaliana (Fig. 3c) 7,8,11,12 . Unlike in the condensin II mutants, the degree of centromere polarization in mutants of NE proteins was low (Fig. 3b-e). This suggests functional redundancy among the SUN and KASH proteins, respectively. Other unidentified proteins may also be involved in CII-LINC-mediated centromeric scattering. For example, centromere distribution is normal in myosin-xi-i, suggesting the existence of other proteins that link actin and WIT, specifically in late mitosis, instead of myosin XI-I. A similar idea was proposed for nuclear lamina-mediated centromere stabilization. At least the known CRWN-interacting proteins, SUN1 and SUN2 (ref. 38 ), are unlikely to be involved in this process (Fig. 4b,c). Moreover, an absence of CRWN1 and CRWN2 does not result in the dissociation of centromeres from the NE during interphase 16 , suggesting the existence of other proteins required for NE association of centromeres in interphase (Fig. 5c). Intriguingly, NUCLEOPORIN1, a component of the nuclear pore complex, interacts directly with pericentromeric chromatin 18,39 and CRWNs 40 , indicating its possible function in interphase centromere distribution. A recent study comprehensively identified putative NE integral proteins, the nuclear factors associated with the NE 41 . Future characterization of known nuclear pore complex components and novel NE-related proteins with respect to centromere association with the NE is needed to fully understand how the position of centromeres is regulated during the cell cycle in A. thaliana.
Previous studies in yeast 42,43 , D. melanogaster 44,45 and human cells 46 have proposed a role for centromere clustering in determining the spatial organization of chromosomes during interphase. However, the significance of adopting a scattered centromere distribution in nuclear events remains unknown. In this study, Hi-C analysis revealed that the defects in both types of centromere positioning, scattering and stabilization, clearly altered genome-wide chromatin organization, including inter-chromosomal interactions (Fig. 6a,b). However, these alterations were not accompanied by widespread changes in gene expression (Supplementary Figs. 10 and 11), which is in agreement with the moderate effects on the transcriptome in another condensin II mutant, cap-d3 (ref. 13 ), and in the crwn1 single mutant (ref. 18 ) that shows similar alterations in chromatin interactions as the crwn1/4 double mutant 18,31 .
These results indicate a robustness of local chromatin organization associated with gene regulation, despite marked changes in centromere distribution.
By contrast, increased DNA damage sensitivity was found to be a common phenotype among mutants defective in centromere distribution (Fig. 7). There are two main explanations for this sensitivity: (i) impaired repair of damaged DNA; or (ii) impaired maintenance of genome stability that is required to avoid damage. Because the gene expression response to DNA damage was not majorly affected in the mutants ( Supplementary Fig. 11), the latter is more likely. In Saccharomyces cerevisiae and D. melanogaster, both having chromosomes in Rabl configuration, destabilization of centromere positioning leads to genome instability, which may be caused by the impaired silencing of transposable elements and repetitive DNA [47][48][49] . However, enhanced expression of transposable elements was not observed in either the cap-h2-2 or crwn1/4 mutant (Supplementary Data 1 and 2), suggesting that there is another cause of genome instability in plants. In addition, our Hi-C analysis suggested the importance of centromere stabilization in the structural organization of pericentromeric regions (Fig. 6e-g). The centromere is considered to be intrinsically fragile, probably because of the high density of repetitive sequences, resulting in its susceptibility to DNA stress 50 . The increased susceptibility to DNA damage at the chromocentre in the crwn1/4 mutant led us to speculate that the aberrant organization of pericentromeric chromatin makes centromeres more vulnerable. However, this hypothesis is not applicable to the impaired integrity of centromeric DNA in cap-h2-2, because the organization of pericentromeric chromatin is essentially like that of the wild type. Alternatively, there may be differences in chromatin conformation, but this would need a higher resolution Hi-C analysis. In conclusion, centromere distribution is possibly linked to genome stability but does not affect gene regulation in A. thaliana.

Generation of transgenic plants. Plants expressing pCAP-H2::CAP-H2-GFP
Confocal imaging. One-week-old plants were used for confocal imaging, unless stated otherwise. For time-lapse imaging of CAP-G2-GFP and the imaging of propidium iodide-stained roots, the roots were observed under an FV1200 inverted laser confocal microscope equipped with a GaAsP detector (Olympus) using a 473 nm LD laser for GFP and a 559 nm LD laser for propidium iodide. To image CAP-G2-GFP, CAP-H2-GFP, CAP-D3-GFP, tdTomato-CENH3, VENUS-CENH3, CTC1-tdTomato and H2B-GFP, the roots and leaves of the samples were observed under an inverted fluorescence microscope (IX81; Olympus), which included a laser (488 nm for GFP and 561 nm for tdTomato detection) equipped with a confocal scanning unit (CSU-X1; Yokogawa) and an Andor Neo 5.5 sCMOS camera (Oxford Instruments). The z-stacks were reconstructed into a maximum projection view using the ImageJ software. The trajectories of centromeres were analysed using the ImageJ software plugin MTrackJ (https://imagescience.org/ meijering/software/mtrackj/). All imaging analyses were repeated independently at least twice with similar results.
Co-IP assays. The coding sequences without the stop codon of CAP-G2, CENH3, SUN1 and SUN2 were amplified from A. thaliana cDNA. The clones of CAP-G2 and CENH3 were subcloned into the pENTR-D/TOPO vector, and SUN1 and SUN2 were subcloned into the pDonr201 vector, according to the manufacturer's protocol. The primers used for cloning are listed in Supplementary Table 1. By LR recombination with LR clonase II, the fragment of CAP-G2 was transferred to pMDC43 (ref. 57 ) and the fragments of SUN1, SUN2 and CENH3 were transferred to pGWB541 or pGWB560 to construct Gateway destination vectors harbouring the target protein fused to GFP, YFP or a monomeric red fluorescent protein, tagRFP, driven by the CaMV p35S promoter. The final plasmids were mobilized into A. tumefaciens (strain GV3101 pMP90).
The fluorescent fusion proteins were transiently expressed in N. benthamiana leaves by Agrobacterium-mediated infiltration, as described previously 14 . Leaves were harvested 4 or 5 d after inoculation. Immunoprecipitation was performed using a µMACS GFP Isolation Kit (Miltenyi Biotec). Leaves (1.0-2.0 g) were homogenized in two volumes of µMACS lysis buffer containing a protease inhibitor cocktail for plant cell and tissue extracts (Sigma-Aldrich). The lysate was then filtered through two layers of Miracloth (Merck), mixed with anti-GFP antibody-conjugated magnetic beads and incubated at 4 °C for 30 min. The GFP-fusion proteins were purified using a magnetic column according to the manufacturer's protocol. Anti-GFP antibody (ab290; Abcam) (1:1,000) and anti-tagRFP antibody (catalogue number: R10367; Thermo Fisher Scientific) (1:500) were used as primary antibodies. A horseradish peroxidase (HRP)-conjugated anti-rabbit IgG pAb-HRP (458, MBL) (1:10,000) was used as the secondary antibody. Chemiluminescence from the target proteins of each antibody was visualized using ImmunoStar LD (Wako) on a Fusion Pulse system (Vilber Lourmat). This experiment was independently repeated twice with similar results.
Yeast two-hybrid assay. The vectors, pENTR-D/TOPO harbouring CAP-H2 or GAP-G2 and pDonr201 harbouring SUN1 or SUN2 were prepared as described above. Using LR recombination with LR clonase II (Invitrogen), the fragments were transferred to pDEST_GADT7 or pDEST_GBKT7. The Y2HGold yeast strain (Takara Bio) was transformed using the Frozen-EZ Yeast Transformation II kit (Zymo Research), according to the manufacturer's protocol. Transformants were selected on SD/-Leu/-Trp medium. Protein interactions were analysed on SD/-Leu/-Trp/-His/-Ade medium. pGADT7 and pGBKT7 were used as negative controls. This experiment was independently repeated twice with similar results.

DNA damage treatment.
For the DNA damage sensitivity test, 5-d-old seedlings preincubated on vertically oriented MGRL plates were transferred to plates containing various concentrations of zeocin (Invitrogen), and their primary root tip positions were marked. After 2 and 4 d of incubation, the lengths of the newly elongated primary roots from the marked positions were determined using ImageJ software (http://rsb.info.nih.gov/ij/). This experiment was independently repeated twice with similar results.
For RNA-seq analysis, 14-d-old seedlings preincubated on vertically oriented MGRL plates were transferred to a container containing a half-strength MGRL solution without sucrose. After 3 h of acclimation to hydroponic conditions, the seedlings were transferred to different containers containing a half-strength MGRL solution with 10 μM zeocin. Whole roots were harvested after treatment with zeocin for 0, 3 and 48 h, and subsequently subjected to total RNA extraction and RNA-seq analysis.
Comet assay. The plants treated with or without zeocin for 2 d were subjected to a comet assay. To extract nuclei, root tips (1 cm) of 100 plants were minced in PBS with a razor blade approximately 100 times, followed by filtration through a CellTrics 30-µm filter (Sysmex Partec) with centrifugation at 800g for 1 min. The comet assay with the N/N method was performed using a CometAssay reagent kit for single cell gel electrophoresis assay (Trevigen), as described previously 60 . First, 10 µl of the nuclear extract was mixed with 100 µl of molten LMAgarose held at 37 °C, and 50 µl of the mixture was immediately pipetted onto CometSlide. The slides were kept at 4 °C in the dark for 30 min, then immersed in a lysis solution (2.5 M NaCl, 10 mM Tris-HCl, pH 8.0, 100 mM EDTA (self-prepared)) for 20 min at room temperature. Slides were washed three times with Tris-borate EDTA buffer for 5 min on ice. Slides were subjected to electrophoresis on ice with Tris-borate EDTA buffer 1 V cm −1 for 6 min. The slides were immersed in 1% Triton-X100/PBS for 10 min at room temperature and then washed twice with 70% ethanol for 5 min and twice with 96% ethanol for 5 min. Slides were dried at 37 °C for 1 h. For staining the nuclei, 100 µl of SYBR Gold Nucleic Acid Gel Stain (10,000× concentrate in dimethylsulfoxide; Thermo Fisher Scientific) diluted 30,000 times with Tris-EDTA buffer pH 8.0 was placed onto CometSlide and kept for 30 min at room temperature in the dark. Slices were rinsed with water and dried completely at 37 °C. Images of nuclei were obtained by inverted fluorescence microscopy on a BX53 microscope (Olympus) equipped with a DOC-CAM HR CCD camera (Molecular Devices). Images were analysed using the ImageJ software plugin CometAssay distributed by the University of North Carolina School of Medicine, USA (https://www.med.unc.edu/microscopy/resources/ imagej-plugins-and-macros/comet-assay). This experiment was independently repeated twice with similar results.
Immunostaining of γH2AX. Plants treated with or without zeocin for 2 d were subjected to γH2AX immunostaining. Immunostaining of γH2AX was performed as described previously 33 . The root tips of seedlings treated with or without zeocin were fixed with 4% paraformaldehyde/PBS for 40 min and then washed in PBS for 5 min at 20 °C. Washed seedlings were incubated in digestion buffer, which is composed of 1% driselase (Sigma-Aldrich), 0.5% cellulose R-10 (Wako) and 0.025% pectolyase Y23 (Wako) in water, for 45 s at 37 °C. After three washes in PBS for 5 min each, the root tips were placed on a MAS-01 adhesive glass slide (Matsunami) and squashed using a cover glass. The slides were immediately placed in liquid N 2 and the cover glass was removed. After the slides were dried, they were immersed in 0.5% (v/v) Triton-X100/PBS for 15 min and then washed three times in PBS for 5 min each at 20 °C. Four per cent (w/v) bovine serum albumin (BSA)/ PBS was dropped onto the sample area. After 30 min, the 4% BSA/PBS solution was removed from the slide and then a mixture of primary antibody was added. The slides were incubated in a humid box at 4 °C overnight and then washed in PBS for 5 min. Next, a mixture of secondary antibody and 1% (w/v) BSA/PBS was added. The slides were incubated at 37 °C for 60 min in the dark and then washed twice in PBS for 5 min. Rabbit anti-γH2AX 33 was used as the primary antibody (1:100) and anti-rabbit IgG Alexa Fluor 488 (Thermo Fisher Scientific) as the secondary antibody (1:1,000) in 1% BSA/PBS solution. Finally, slides were immersed in 0.1 μg ml −1 DAPI (Roche)/PBS for 5 min and then mounted under a cover glass with VECTASHIELD (Vector Laboratories). The specimens were observed using a CSU-X1 based confocal system. This experiment was independently repeated twice with similar results.
Hi-C sample preparation. Aerial parts (~2.0 g) of 2-week-old seedlings were used for Hi-C sample preparation. Nuclei isolation, digestion with HindIII, fill-in biotinylation, ligation and purification of the Hi-C samples were performed as described previously 61 . For preparation of the sequencing library, the purified Hi-C sample (500 ng) was diluted to 500 μl with distilled water, and 500 μl of 2× binding buffer (10 mM Tris, 1 mM EDTA, 2 M NaCl) was added. The diluted Hi-C samples were fragmented to a mean size of 300 bp by sonication using a Covaris M220 sonication system in a milliTUBE 1 ml AFA Fibre (Covaris). The parameters of the programme were as follows: power mode, frequency sweeping; time, 20 min; duty cycle, 5%; intensity, 4; cycles per burst, 200; temperature (water bath), 6 °C. Biotin-labelled Hi-C samples were then enriched using MyOne Streptavidin C1 magnetic beads (Veritas). For this, 60 μl of streptavidin beads were washed twice with 400 μl of Tween wash buffer (5 mM Tris, 0.5 mM EDTA, 1 M NaCl, 0.05% Tween-20). Recovery of streptavidin beads was performed by placing the tubes on a magnetic stand. Subsequently, the beads were added to 1 ml of sheared Hi-C sample. After 15 min of incubation at room temperature under rotation, the supernatant was removed, and the beads binding biotinylated Hi-C fragments were resuspended in 400 μl of 1× binding buffer. Beads were then washed once in 60 μl of resuspension buffer (RSB; Illumina), and finally resuspended in 50 μl of RSB. The enriched biotinylated DNA fragments were subjected to library construction on beads using the KAPA HyperPrep Kit for Illumina (Roche) according to the manufacturer's protocol, with 18 cycles of PCR for library amplification. The amplified DNA fraction (50 μl) was corrected and purified using Agencourt AMPure XP (Beckman Coulter), following the standard protocol, and resuspended in 15 μl RSB. The quality of the Hi-C libraries was assessed using the TapeStation with High Sensitivity D1000 ScreenTape (Agilent Technologies). The libraries were then sequenced on a NextSeq 500 sequencer (Illumina) in the 150 bp paired-end mode.
Hi-C data analysis. Raw Hi-C sequencing reads were trimmed to 30 bp and aligned to a TAIR10 genome using subread-align 62 v.2.0.0 with the parameters: -u (unique only), -I 0 (no indels), -M 0 (no mismatches). Subsequently, the aligned Hi-C data were further processed with the help of HiCdat 63 and contact matrices of 50 kb and 100 kb bin sizes were generated (see Supplementary Table 2 for number of valid contacts). The minimal distance to filter inward pairs was set to 1 kb and to filter outward-facing pairs to 25 kb. Pairs fulfilling these criteria have not been included in the Hi-C matrices. Distance-normalization was performed while loading the data using the f.load.samples() function including 25 repetitions in the R package HiCdatR 63 . IDE calculation, differential analysis, correlation analysis between samples and analysis of the first principal component were also performed using HiCdatR. Principal component analyses were performed on entire chromosome lengths using a 100 kb bin size. Genomic coordinates of pericentromeric regions to calculate IDEs were defined as follows: Chr1, 10-18 Mb; Chr2, 1-8 Mb; Chr3, 10-17 Mb; Chr4, 1.5-6.5 Mb; Chr5, 9-16 Mb. Normalized differences (Fig. 6g) were calculated as follows: for each individual chromosome, cis-contacts of Hi-C matrices were initially normalized to counts per million. Subsequently, the raw difference between samples was determined (mutant minus wild type), and the raw difference was divided by the mean contact strength between the mutant and the wild type ((norm WT + norm mutant)/2). The DLRs were calculated on counts per million-normalized Hi-C matrices, using the following parameters: local, radius of 3 Mb around the region of interest. Distant contacts were collected between 3 and 12 Mb. Distant contacts were sampled only towards the end of the chromosome, where the region of interest was located ≥12 Mb away from the telomere.

Microarray analysis.
Total RNA was extracted from whole roots of seedlings using PureLink Plant RNA Reagent (Thermo Fisher Scientific), according to the manufacturer's protocol. Microarray analysis was performed using an Agilent Arabidopsis microarray platform as previously described 64 . Robust Multichip Average normalization was performed for signals of microarray probes using the limma package 65 in R (v.2.12.1; R Core Team). The statistical significance of differences in gene expression was determined using a Student's t-test and a Benjamini-Hochberg corrected false discovery rate 66 .
RNA-seq sample preparation. Whole roots (~0.1 g) of 2-week-old seedlings treated with or without zeocin were used for RNA-seq sample preparation. Total RNA was extracted using the Monarch Total RNA Miniprep Kit (New England Biolabs), according to the manufacturer's protocol. The extracted total RNA was qualified using TapeStation with RNA ScreenTape (Agilent Technologies). Then, 500 ng of the extracted RNA was subjected to RNA-seq library preparation using the NEBNext Ultra II RNA Library Prep Kit for Illumina (New England Biolabs), according to the manufacturer's protocol. The amplified cDNA fraction (25 μl) was corrected and purified using Agencourt AMPure XP (Beckman Coulter), following the standard protocol and resuspended in 21 μl RSB. The quality of the RNA-seq libraries was assessed using a TapeStation with D1000 ScreenTape (Agilent Technologies). The libraries were sequenced on a NextSeq 500 sequencer (Illumina) in the 75 bp single-end mode.
RNA-seq data analysis. The reads were first quality checked with FastQC v0.11.9 and then trimmed with Trimmomatic 67 v.0.39 using MINLEN:60 HEADCROP:5 CROP:50. STAR 68 v.2.7.10a with basic options was used to map the trimmed reads to a modified version of Araport11 downloaded from https://plants. ensembl.org (see Supplementary Table 3 for number of mapped reads). Because CRWN4 (AT5G65770) is co-annotated with the same features as a gene model for ATBCAT-5 (AT5G5780.2), we deleted the latter from our annotation file to ensure that reads were properly mapped to the CRWN4 locus. The mapped reads were assigned to genomic features using featureCounts 69 v.2.0.0 with default settings. The assembled raw count matrices were imported into R and analysed using the DESeq2 package 70 v.1.34.0. Genes with a sum of reads <10 across all samples were removed before principal component and differential expression analyses. EnhancedVolcano v.1.13.2 was used to plot DEGs with a false discovery rate <0.1 and log 2 (fold change) ≥1.
Bisulfite-sequence sample preparation. Aerial parts (~0.1 g) of 2-week-old seedlings were used for bisulfite-sequence (BS-seq) sample preparation. Genomic DNA was extracted using the DNeasy Plant Mini Kit (Qiagen) according to the manufacturer's instructions. The integrity of gDNA was checked by agarose gel electrophoresis; the DNA was quantified using Quant-IT PicoGreen (Invitrogen). BS-seq libraries were prepared using the Accel-NGS Methyl-Seq DNA library kit (Swift BioSciences). Briefly, 200 ng of gDNA were fragmented using a Covaris LE220 focused-ultrasonicator to a target peak size of 400-700 bp. The fragmented DNA was then subjected to bisulfite conversion using the EZ DNA Methylation-Gold Kit (Zymo Research), following the manufacturer's instructions. The bisulfite-treated single-strand DNA fragments were repaired and truncated adaptors 1 and 2 were ligated to the 3′-and 5′-ends of the fragments, respectively. The truncated adaptor-ligated DNA was amplified with an indexed primer to complete the BS-seq libraries using a full-length adaptor. The final libraries were then quantified using quantitative PCR, according to the PCR Quantification Protocol Guide (KAPA Library Quantification kits for Illumina Sequencing platforms, Roche), and qualified using TapeStation with D1000 ScreenTape (Agilent Technologies). The libraries were then sequenced using a NovaSeq 6000 sequencer (Illumina).

BS-seq data analysis.
Raw sequencing reads were aligned to a modified TAIR10 genome using bismark 71 v.0.23.0 with the following parameters: -q --score-min L,0,-0.2 --ignore-quals --no-mixed --no-discordant --dovetail --maxins 500. Subsequently, the methylation scores in CG, CHH and CHG contexts (H being any base except G) were assessed using the methylKit 72 R package v.1.22.0, with a minimal coverage of 10 and minimal quality score of 20. Methylation differences were scored using the function calculateDiffMeth(). Overall methylation levels and plots were determined using custom R scripts, and only methylation levels above a threshold were considered (CG = 40%, CHG = 20% and CHH = 10%). Circos plots were generated using the R package circlize 73

Statistics
For all statistical analyses, confirm that the following items are present in the figure legend, table legend, main text, or Methods section.

n/a Confirmed
The exact sample size (n) for each experimental group/condition, given as a discrete number and unit of measurement A statement on whether measurements were taken from distinct samples or whether the same sample was measured repeatedly The statistical test(s) used AND whether they are one-or two-sided Only common tests should be described solely by name; describe more complex techniques in the Methods section.
A description of all covariates tested A description of any assumptions or corrections, such as tests of normality and adjustment for multiple comparisons A full description of the statistical parameters including central tendency (e.g. means) or other basic estimates (e.g. regression coefficient) AND variation (e.g. standard deviation) or associated estimates of uncertainty (e.g. confidence intervals) For null hypothesis testing, the test statistic (e.g. F, t, r) with confidence intervals, effect sizes, degrees of freedom and P value noted

Software and code
Policy information about availability of computer code Data collection We used ImageJ software for the collection of imaging data, comet assay data, and data of root phenotype. NGS datasets were generated on Illumina's sequencers.
For manuscripts utilizing custom algorithms or software that are central to the research but not yet described in published literature, software must be made available to editors and reviewers. We strongly encourage code deposition in a community repository (e.g. GitHub). See the Nature Portfolio guidelines for submitting code & software for further information.

March 2021
Data Policy information about availability of data All manuscripts must include a data availability statement. This statement should provide the following information, where applicable: -Accession codes, unique identifiers, or web links for publicly available datasets -A description of any restrictions on data availability -For clinical datasets or third party data, please ensure that the statement adheres to our policy The data supporting the findings of this study are available within the paper and its supplementary information files. The microarray data (GEO ID: GSE179466) are available on the GEO website (https://www.ncbi.nlm.nih.gov/geo/). The Hi-C (DRA013016), RNA-seq (DRA014243), BS-seq (DRA014250) data are available on the DDBJ website (https://www.ddbj.nig.ac.jp/index-e.html). We used a modified version of Araport11 downloaded from https://plants.ensembl.org for RNA-seq data analysis and a TAIR10 genome downloaded from https://www.arabidopsis.org/index.jsp for Hi-C and BS-seq data analysis.

Field-specific reporting
Please select the one below that is the best fit for your research. If you are not sure, read the appropriate sections before making your selection.

Life sciences Behavioural & social sciences Ecological, evolutionary & environmental sciences
For a reference copy of the document with all sections, see nature.com/documents/nr-reporting-summary-flat.pdf

Life sciences study design
All studies must disclose on these points even when the disclosure is negative.

Sample size
No sample-size calculation was performed. Samples size was chosen according to convention in the filed of plant cell biology for all experiments.
Data exclusions No data were excluded.

Replication
All data shown is based on multiple biologically independent replicates. The number of replicates for each experiment was stated in the Figure or