The Arabidopsis condensin CAP-D subunits arrange interphase chromatin

Condensins are best known for their role in shaping chromosomes. However, other functions as organizing interphase chromatin and transcriptional control have been reported in yeasts and animals. Yeasts encode one condensin complex, while higher eukaryotes have two of them (condensin I and II). Both, condensin I and II, are conserved in Arabidopsis thaliana, but so far little is known about their function. Here we show that the A. thaliana CAP-D2 (condensin I) and CAP-D3 (condensin II) subunits are highly expressed in mitotically active tissues. In silico and pull-down experiments indicate that both CAP-D proteins interact with the other condensin I and II subunits. Our data suggest that the expression, localization and composition of the condensin complexes in A. thaliana are similar as in other higher eukaryotes. Previous experiments showed that the lack of A. thaliana CAP-D3 leads to centromere association during interphase. To study the function of CAP-D3 in chromatin organization more in detail we compared the nuclear distribution of rDNA, of centromeric chromocenters and of different epigenetic marks, as well as the nuclear size between wild-type and cap-d3 mutants. In these mutants an association of heterochromatic sequences occurs, but nuclear size and the general methylation and acetylation patterns remain unchanged. In addition, transcriptome analyses revealed a moderate influence of CAP-D3 on general transcription, but a stronger one on transcription of stress-related genes. We propose a model for the CAP-D3 function during interphase, where CAP-D3 localizes in euchromatin loops to stiff them, and consequently separates centromeric regions and 45S rDNA repeats.


INTRODUCTION
The spatial genome arrangement is important to regulate the access of proteins to DNA (Gibcus and Dekker, 2013), since the folding of chromatin allows or impedes interactions between distinct loci and their regulatory sequences (Doğan and Liu, 2018;Robson et al., 2019;Stam et al., 2019;Szabo et al., 2019). Thus, a better knowledge of the nuclear organization during interphase could help to understand processes like replication, DNA repair, recombination and transcription. During interphase, in A. thaliana (Pecinka et al., 2004) as well as in other higher eukaryotes, the chromosomes occupy discrete regions called chromosome territories. Although this higher order chromatin arrangements is highly conserved, species-specific structural and functional features of nuclear organization exist in metazoans and protists (Cremer and Cremer, 2010;Cremer et al., 2018). In contrast to mammals (Boyle et al., 2001;Mayer et al., 2005) and birds (Habermann et al., 2001), A. thaliana chromosome territories prefer no particular position within the nucleus (Pecinka et al., 2004). Chromocenters are chromatin structures intensely stained by DNA-specific dyes and represent condensed heterochromatin regions in interphase nuclei (Jost et al., 2012). In A. thaliana chromocenters incarnate centromeric and pericentromeric heterochromatin located near the nuclear periphery and the nucleolus Schubert et al., 2012). For the maintenance of chromocenters different proteins related to methylation, ATPases and nuclear periphery components were described in A. thaliana (Soppe et al., 2002;Moissiard et al., 2012;Wang et al., 2013;Poulet et al., 2017). To explain the organization of chromosome territories in A. thaliana interphase nuclei, a rosette model was proposed . Based on cytological observation and later support by computer simulations (de Nooijer et al., 2009) and Hi-C data (Feng et al., 2014;Liu et al., 2016), this model assumes that the chromosomes are organized as chromatin loops emanating from the chromocenters. Structural Maintenance of Chromosomes (SMC) complexes are present in prokaryotes and eukaryotes (Cobbe and Heck, 2004). They are essential for chromatin organization and dynamics, gene regulation and DNA repair. In eukaryotes six conserved SMC subunits form the core of three different complexes: cohesin, involved in sister chromatids cohesion and interphase chromatin arrangement; condensin, involved in mitotic and meiotic chromosome organization (van Ruiten and Rowland, 2018;Skibbens, 2019); and the SMC5/SMC6 complex, mainly involved in DNA repair and replication (Jeppsson et al., 2014). Animals have two condensin complexes, condensin I and II (Ono et al., 2003). In yeasts, only one condensin complex analogous to animal condensin I is present (Freeman et al., 2000;Hirano, 2012a). Condensin I and II share a core formed by SMC2 and SMC4 and differ in the associated proteins, which are in condensin I CAP-H, CAP-D2 and CAP-G, and in condensin II CAP-H2, CAP-D3 and CAP-G2 (Ono et al., 2003;Hirano, 2012a). This composition is conserved in higher eukaryotes, although in Drosophila the subunit CAP-G2 of condensin II has not been detected (Herzog et al., 2013). As proposed for A. thaliana (Fig. 1), plants apparently have condensin I and II. Condensins have been widely studied in human, animals and yeast for their role in shaping chromosomes. Together with topoisomerase II condensins form a scaffold within human somatic metaphase chromatids (Maeshima and Laemmli, 2003). Depletion of condensin I causes short fuzzy metaphase chromosomes while the depletion of condensin II causes long and curly chromosomes (Ono et al., 2003;Green et al., 2012). Besides aberrant chromosome morphologies, chromosomes lacking several condensin subunits show anaphase bridges and other segregation defects (Freeman et al., 2000;Hudson et al., 2003;Ono et al., 2003Ono et al., , 2004Hirota et al., 2004;Savvidou et al., 2005;Gerlich et al., 2006;Hartl et al., 2008). Both complexes may form DNA loops resulting in chromosome compaction (Elbatsh et al., 2019;Gibcus et al., 2018;van Ruiten and Rowland, 2018;Walther et al., 2018). Condensin I and II complexes show a distinct subcellular localization during the mammalian cell cycle. In human and rat during interphase, condensin I occurs in the cytoplasm, condensin II in the nucleus (Hirota et al., 2004;Ono et al., 2004). During mitosis, condensin I and II localize along the chromosome arms in an alternate fashion, and both are enriched at the centromeres (Ono et al., 2003;Ono et al., 2004;Savvidou et al., 2005). In addition to the canonical role in metaphase chromosome formation, condensins are also involved in gene expression and chromatin organization during interphase (Wallace and Bosco, 2013;Wallace et al., 2015). In mouse and human, condensin II localizes at the promoters of active genes and is required for normal gene expression (Dowen et al., 2013;Yuen et al., 2017;). In Drosophila, CAP-D3 together with the RetinoBlastoma protein RBF1, regulates gene clusters involved in tissue-specific functions (Longworth et al., 2012), and condensin II promotes the formation of chromosome territories and keeps repetitive sequence clusters apart from each other (Hartl et al., 2008;Bauer et al., 2012;Hirano, 2012b;Rosin et al., 2018). Also A. thaliana posseses the components for both condensin complexes (Schubert, 2009;Smith et al., 2014). In contrast to other organisms, A. thaliana has two SMC2 homologs, SMC2A and SMC2B with redundant functions (Siddiqui et al., 2003) (Fig. 1). As in other species, SMC4, CAP-H and CAP-H2 are present within chromosomes and are required for normal metaphase chromosome compaction (Fujimoto et al., 2005;Smith et al., 2014). During interphase, CAP-H is present in the cytoplasm of protoplasts while the condensin II subunits CAP-H2 and CAP-D3 were detected in the nucleolus and euchromatin, respectively (Fujimoto et al., 2005;Schubert et al., 2013). CAP-D2 and CAP-D3 prevent the association of centromeres and induce chromatin compaction (Schubert et al., 2013). The requirement of the condensin II-specific subunits CAP-H2 and CAP-G2 for keeping centromeres apart has been confirmed by Sakamoto et al. (2019). In addition, these authors showed that condensin II is necessary for the correct spatial arrangement between centromeres and rDNA arrays. Condensins are highly conserved, but have not been studied extensively in plants. Here we analyze the A. thaliana CAP-D2 and CAP-D3 condensin subunit expression patterns, their cellular localization and interaction with other condensin subunits and additional proteins for better understanding their functions. We demonstrate that also A. thaliana forms specific condensin I and II complexes, and show that CAP-D3 mediates the spatial separation of chromocenters, without altering the global methylation pattern and nuclear ultrastructure. Finally, we suggest a model explaining the action of CAP-D3 to prevent the association of chromocenters.

CAP-D2 and CAP-D3 are highly expressed in meristematic tissues
Based on in silico analysis using the Arabidopsis eFP Browser (Winter et al., 2007), A. thaliana CAP-D2 (At3g57060) and CAP-D3 (At4g15890) have a similar expression pattern. Both proteins are highly expressed in the shoot apex, roots, flower buds and vegetative rosette leaves. Their expression is lower in cotyledons, rosette leaves after bolting, mature flowers, siliques and embryos ( Figure S1). To corroborate the in silico data we assessed the transcription of both genes in seedlings, mature rosette leaves, roots and flower buds by quantitative real-time RT-PCR. The highest transcription of both genes was observed in flower buds, the lowest in seedlings. The transcription level of CAP-D2 is 25.6, 14.8 and 3.5 times higher in flower buds, roots and leaves, respectively, than in seedlings. Similarly, the CAP-D3 transcription is 18.3, 9.4 and 4.4 times higher in flower buds, roots and leaves respectively, than in seedlings (Fig. 2). The activity of the CAP-D2 and CAP-D3 promoters was evaluated in A. thaliana transgenic lines expressing different versions of the promoters fused to the -glucuronidase (GUS) reporter gene (Fig.  3a). Six presumed promoters of different length were analyzed for CAP-D2 and two for CAP-D3. The promoter region of CAP-D2 contains two putative E2F binding sites at -345 bp and -114 bp upstream from the start of the gene (Schubert et al., 2013). Two promoter lengths were analyzed: a promoter that comprises 1156 bp upstream of the start of CAP-D2 (Pro4), and a short promoter of 391 bp (Pro7). In addition, promoter proximal introns can enhance the expression of a gene by a mechanism known as Intron-Mediated Enhancement (IME) (Rose et al., 2008). The putative enhancing ability of CAP-D2 introns was analyzed in silico with the web tool IMEter (Parra et al., 2011). The IMEter score is positively correlated to the enhancing ability of an intron. For CAP-D2 the two first introns have positive IMEter scores of 12.13 and 2.36, respectively. Thus, it is likely that they enhance expression. These introns were included in the analysis in combination with the long and short promoters of CAP-D2: Pro5 (long promoter) and Pro8 (short promoter) include Intron1; and Pro6 (long promoter) and Pro9 (short promoter) include both Intron1 and Intron2. The promoter region of CAP-D3 contains also two putative E2F binding sites at -397 bp and -84 bp (Schubert et al., 2013). However, the IMEter scores of the first two CAP-D3 introns were negative, -13.20 and -5.88 respectively. Thus, it is unlikely that they enhance expression. Therefore, for CAP-D3 the introns were not considered and only a long promoter at -1318 bp (Pro10) and a short promoter at -474 bp (Pro11) from the start of the gene were analyzed. T1 transgenic plants with the different versions of CAP-D2 and CAP-D3 promoters were stained for GUS analysis (Fig. 3b). Only for Pro4 no positive plants could be isolated. The CAP-D2 promoter version Pro5 (n=7) was active in stipules (small organs at the base of the leaves), leaf vascular tissue and root tip meristems. Pro6 (n=6) had weak activity in root tips. All Pro7 plants (n=21) showed GUSstaining in leaf vascular tissue and root tips, and 16 plants also in stipules. All Pro8 plants (n=23) presented GUS activity in the apical meristem and root tips, and 16 of them also in leaf vascular tissue. Pro9 (n=5) showed activity in roots, and 3 plants also weakly in the apical meristem (Fig. 1d). Therefore, all CAP-D2 promoter versions were active in root tips, but the staining was stronger in the short promoter versions (Pro7, Pro8 and Pro9) than in the long ones (Pro5 and Pro6). In addition, the CAP-D2 short promoters showed an activity in the apical meristem and versions that included the second intron (Pro6 and Pro9) lost the staining in the leaf vascular tissue. CAP-D3 Pro10 showed no activity, and for Pro11 (n=8), the plants showed activity in the apical meristem and root tips. For both, CAP-D2 and CAP-D3, the expression can be driven more effectively by the short promoter, which contains the E2F sites. Taken together quantitative real-time RT-PCR and GUS activity staining demonstrated that, CAP-D2 and CAP-D3 are highly expressed in meristematic tissues (root tip meristem, flower buds, apical meristem) and young leaves and less expressed in mature leaves. The low transcription observed in seedlings could be due to a low amount of meristematic tissue in the sample since just one-week old seedlings were used for RNA isolation.

CAP-D2 and CAP-D3 interact with the other condensin subunits in specific complexes
CAP-D2 and CAP-D3 are specific components of condensin I and II complexes, respectively. The presence of CAP-D2 and CAP-D3 as well as the other condensin complex subunits in A. thaliana was previously confirmed (Smith et al., 2014), but whether the complexes are formed by the same subunits as in non-plant species is unknown. To predict a composition of each complex we identified putative interactors of CAP-D2 ( Figure S2a) and CAP-D3 ( Figure S2b) in silico using the STRING program (http://string-db.org/; Szklarczyk et al., 2019). At the high score of >0.90 the following proteins: SMC2A (At5g62410), SMC2B (At3G47460) and SMC4 (At5g48600) were identified in interaction networks of both CAP-D2 and CAP-D3, while CAP-G (At5g37630) and CAP-H (At2g32590) were found as interactors of CAP-D2, and CAP-G2 (At1g64960) and CAP-H2 (At3g16730) as specific interactors of CAP-D3, respectively. Due to the presence of SMC2A, SMC2B and SMC4 in both interaction networks they may be involved in the formation of condensin I as well as of condensin II. In silico analysis using the STRING program identified besides cohesin subunits also SMC5/6 complex subunits as CAP-D2 and CAP-D3 interacting partners (Zelkowski et al., 2019). To confirm these in silico results and to determine the composition of each complex experimentally, CAP-D2 and CAP-D3 were fused to a GS-tag, and affinity-purified from A. thaliana PSB-D suspension cultured cells ( Figure S3). The proteins co-purifiying with CAP-D2-GS and CAP-D3-GS were identified by mass spectrometry. The putative subunits of the condensin I complex, SMC2A, SMC2B, SMC4, CAP-H and CAP-G, were detected with high scores in the CAP-D2-GS eluates of three affinity purifications performed. Similarly, the putative subunits of the condensin II complex, SMC2A, SMC4, CAP-H2 and CAP-G2, were detected in the three affinity purifications performed for CAP-D3-GS and SMC2B in two of the affinity purifications (Table 1, Fig. 1). Like in the in silico analysis, CAP-H, CAP-G and CAP-H2, CAP-G2 were identified as specific components of the condensin I and condensin II complexes, respectively, while SMC2A, SMC2B and SMC4 coprecipitated with both CAP-D2 and CAP-D3.The results indicate that A. thaliana, similar as mammals, chicken and C. elegans (Hirano, 2012a;Onn et al., 2007), comprises specific condensin I and II complexes. Interestingly, in addition to SMC4, both SMC2A and SMC2B may be involved in the formation of both condensin complexes. Among the proteins which co-purified with CAP-D2 (Table S1), other proteins such as the cohesin complex subunit SMC3 were identified. Additionally, the chromatin remodeling factors CHR17 and CHR19; CUL1, a subunit of the SCF ubiquitin ligase complex; HDC1, a histone deacetylase and ELO3, a histone acetyltransferase from the elongator complex were found. Among the proteins copurifying with CAP-D3 (Table S2) were two nucleosome assembly proteins (NAP); CSN1, a subunit of the COP9 signalosome (CSN); the helicase BRAHMA; ELO3, from the elongator complex, and NERD, involved in DNA methylation. The results indicate that both A. thaliana Cap-D genes are highly conserved, and that the corresponding proteins may act in combination with other condensin complex components, as well as with cohesin and SMC5/6 subunits.

Condensin I subunits are localized within nuclei and cytoplasm
Previously, A. thaliana protoplasts have been used to examine the localization of the condensin subunits CAP-H and CAP-H2 (Fujimoto et al., 2005). Therefore, we expressed transiently the coding region of CAP-D2 fused to EYFP (35S::CAP-D2_EYFPc) in A. thaliana mesophyll protoplasts. To visualize EYFP, the protoplasts were immunolabeled with anti-GFP antibodies. We identified CAP-D2 in the cytoplasm and the DAPI-counterstained nucleus (Fig. 4a). In the cytoplasm GFP-negative, but DAPI-positive round chloroplasts were also visible. The free EYFP of the positive control also localized in the cytoplasm and nucleus. Western blot analysis of CAP-D2_EYFPc transformed protoplasts confirmed that the CAP-D2_EYFP protein was intact, and that the visible localization corresponds to the fusion protein (187 kDa), and not to free EYFP (27 kDa) (Fig. 4b). The condensin I subunits, CAP-H and CAP-G fused to EYFP localized also in the cytoplasm and the nucleus ( Figure  S4). Similarly, transient transformation of N. benthamiana leaves revealed the localization of CAP-D2, CAP-H and CAP-G EYFP-fusion proteins in cytoplasm and nuclei too ( Figure S5). Anti-CAP-D2 Western blot antibodies were generated against a recombinant protein containing the last 501 amino acids of CAP-D2 ( Figure S6). The CAP-D2 antiserum can detect amounts as low as 1 ng of the recombinant protein ( Figure S7). The CAP-D2 antiserum detects the CAP-D2 fusion protein from protoplasts (Fig. 4b), but not the CAP-D2 protein from wild-type leaves (data not shown). This may be due to a lower amount of the target protein in leaves compared to that in protoplasts. In protoplast overexpression of CAP-D2 occurred since the reporter construct is under control of the 35S promoter. In order to localize CAP-D2 and CAP-D3 proteins in planta, A. thaliana wild-type plants were transformed with constructs containing the coding region of either gene fused at its C-terminus to enhanced yellow fluorescence protein (EYFP) under the control of the 35S promoter (35S::CAP-D2_EYFPc and 35S::CAP-D3_EYFPc). In both cases, the detection of the proteins in vivo or by immunolocalization with anti-GFP antibodies (also detecting EYFP) was not possible. The same negative result was obtained by reporter constructs with EYPF fused at the N-terminus (35S::CAP-D2_EYFPn and 35S::CAP-D3_EYFPn).

CAP-D3 organizes chromatin during interphase
The involvement of A. thaliana CAP-D3 in compacting chromosome territories (CT) and keeping centromeres apart at interphase has been previously described by Schubert et al. (2013). In Drosophila, CAP-D3 is also involved in the formation of compact chromosome territories (Hartl et al., 2008). To further study the involvement of CAP-D3 in chromatin organization we used two cap-d3 mutants described previously, Cap-D3 SAIL_826_B06 and Cap-D3 SALK_094776 (Schubert et al., 2013) (Fig. 5a,b). To confirm the centromeric clustering and CT dispersion phenotypes in both mutants, a FISH experiment on flow-sorted 4C nuclei was performed with probes specific for the centromere repeat pAL and the chromosome 1 arm territory bottom (CT1B) (Fig. 5c, d). In addition to the number of centromeric pAL signals per nucleus, the areas of the CT1B signals and the nucleus were measured. The median area size of the CT1B signals was 3.9, 4.7 and 4.7 µm 2 for cap-d3 SAIL, cap-d3 SALK and wild-type, respectively (Fig. 5e). No significant differences were found. Thus, we could not confirm the CT dispersion phenotype of the cap-d3 mutants described in Schubert et al. (2013). In addition, no significant differences were found in the nuclear area size between the cap-d3 mutants and wild-type plants (Fig. 5e). On the other hand, we could confirm the centromere-association phenotype. In both cap-d3 mutants the nuclei showed a lower number of centromeric pAL signal clusters than wild-type (Fig. 5d). Around 80% of the cap-d3 mutant nuclei showed less than six pAL signals, while in wild-type only 12% of nuclei had less than six pAL signals (Fig. 5f). To verify that the mutation in the CAP-D3 gene is indeed responsible for the centromeric clustering, a complementation experiment was carried out. cap-d3 SALK mutant plants were transformed with CAP-D3_EYFPc constructs, containing the coding region of CAP-D3 fused to EYFP under the control of the 35S promoter. The centromeric association phenotype was evaluated in cap-d3 SALK complemented plants by FISH and compared with the cap-d3 SALK mutants and wild-type. Only 15% of the complemented nuclei showed less than six centromeric signal clusters, which is similar as the wild-type association levels (Fig. 5g). This confirms that CAP-D3 is responsible for the centromere association in the mutants. Beside centromeres, in A. thaliana, the 45S and 5S rDNAs are heterochromatin-associated sequences. In nuclei of differentiated cells, 45S rDNA containing nucleolar organizing regions tend to associate, but the 5S rDNA loci are often separated (Berr and Schubert, 2007). To examine whether CAP-D3 affects in general the organization of heterochromatin, the distribution of the 45S and 5S rDNA loci was analyzed by FISH in both cap-d3 mutants (Fig. 6a). The majority of 45S rDNA signals is shifted from three signals in wild-type to two signals in the mutants (Fig. 6b). No difference was observed with regard to 5S rDNA since over 70% of the nuclei showed between six and ten signals in the cap-d3 mutant and wild-type plants (Fig. 6c). Thus, the cap-d3 mutants present a higher association of the chromosomal 45S rDNA regions than wild-type, but the number of 5S rDNA signals remains unaffected. A. thaliana centromeres are positioned at the nuclear periphery Fang and Spector, 2005). To test whether the centromere position is influenced by the cap-d3 mutations nuclei were embedded in acrylamide to preserve their 3D structure followed by FISH (3D-FISH) with the centromeric pAL repeats. For each genotype, cap-d3 SAIL, cap-d3 SALK and wild-type, 10 nuclei were analyzed. Optical sections (3D-SIM stacks) were acquired for each nucleus, and the centromere positions were analyzed in the ZEN software tool 'ortho view' (Fig. 6d). In all the cases, the centromeres were localized at the periphery of the nucleus, even when centromere clustering was present in the cap-d3 mutants. Consequently, no deviation in peripheral centromere positioning in wild-type and the cap-d3 mutants was observed.
CAP-D3 does not effect the nuclear distribution of histone marks DNA can be methylated at cytosine as 5-methyl-cytosine (5mC). The methylation of DNA is associated with heterochromatin formation and consequently, it has been found in the chromocenters of A. thaliana . Mouse embryonic stem cells depleted of condensin show a reduction of 5mC (Fazzio and Panning, 2010). In order to test whether such an effect can also be observed in plants, the distribution of methylated DNA in cap-d3 mutants was compared to wild-type by immunodetection of 5mC-specific sites. In both cap-d3 mutants and wild-type the 5mC signals were similarly chromocenter-localized (Fig. 7a). The A. thaliana centromeric repeats are highly methylated in a CpG context (Martinez-Zapater et al., 1986). The use of methylation sensitive enzymes and Southern blot hybridization allowed a more precise determination of the relative DNA methylation level of the centromeric repeats. HpaII and its isoschizomer MspI cleave the same CCGG sequence, but HpaII is methylation sensitive while MspI is not. In wild-type, the centromeric repeats are highly methylated and are thus digestible by MspI (Fig. 7b). The ladder-like pattern corresponds to the monomer, dimer, trimer and higher orders of centromeric repeats. As expected, HpaII does not cut in wild-type DNA. In both cap-d3 mutants, the hybridization pattern is similar to wild-type. Thus, the relative level of CCGG methylation is not altered in the cap-d3 mutants (Fig. 7b). CAP-D3 prevents the clustering of heterochromatin, but the CAP-D3 protein itself localizes in euchromatic regions during interphase. Both types of chromatin are characterized by specific posttranslational histone modifications marks (Fuchs et al., 2006). To evaluate a possible functional association between histone modifications and CAP-D3 functions, the global distribution patterns of different histone marks were compared between the cap-d3 mutants and wild-type. Specific marks for heterochromatin (histone H3K9me1, H3K9me2) and euchromatin (histone H3K4me3, H3K27me3) were tested by indirect immunostaining. In addition, the H3 acetylation marks H3K9ac, H3K14ac, and H3K18ac as well as H3K9+14+18+23+27ac were evaluated. Histone acetylation relaxes chromatin allowing different protein complexes to access DNA (Wang et al., 2014). Thus, histone acetylation is associated with transcription, and hypoacetylation with transcriptional repression. In flow-sorted 4C wild-type nuclei, H3K4me3 localizes in euchromatin and it is absent from chromocenters and the nucleolus. In cap-d3 mutants the localization is identical. H3K9me1 is a heterochromatin-specific histone modification that localizes in the chromocenters in both cap-d3 mutants and wild-type. Finally, the acetylation mark H3K14ac localizes mainly in euchromatin (transcriptionally active chromatin) of wild-type nuclei, but also in the mutants (Fig. 7c). The other histone modifications tested (H3K27me3, H3K9me2, H3K9ac, H3K18ac and H3K14+18+23+27ac) followed also the same pattern in wild-type and the cap-d3 mutants ( Figure S8). Thus, we did not detect obvious differences in the (sub-)nuclear distribution patterns of the different histone marks between wild-type and the cap-d3 mutants.

CAP-D3 moderately affects transcription
To assess if the increased clustering of the centromeric interphase chromatin in the cap-d3 mutants affects gene transcription, the transcriptome of both cap-d3 mutants was compared to wild-type. RNA-sequencing was performed in 4 samples (pooled 4 weeks-old plantlets) for each genotype. After differential expression analysis, we could observe alterations between the cap-d3 mutants and wildtype transcriptomes. The genes with at least 2-fold change transcription and a pAdj ≤ 0.05 were considered as differentially expressed genes (DEG) between two genotypes (Fig. 8a). The smallest difference was observed between cap-d3 SAIL vs. cap-d3 SALK (74 DEG), and the highest between cap-d3 SAIL vs. wild-type (398 DEG). cap-d3 SALK vs. wild-type was intermediate (97 DEG) (Fig.  8b). Both cap-d3 mutants show centromere and 45S rDNA clustering, but cap-d3 SAIL plants showed additional growth defects that are absent in cap-d3 SALK plants. To separate the individual effect of each allele, in further analysis only the DEG shared by both mutants when compared to wild-type were considered. These 83 genes, common to the cap-d3 mutation independently of the specific alleles, are subsequently referred to as "cap-d3 DEG" (Fig. 8b and Table S3). These genes are distributed along all chromosome arms (Fig. 8c). According to their Gene Ontology (GO) enrichment, the cap-d3 DEGs are mainly involved in transcription, particularly in biological processes affecting the response to water, stimuli and stress (Table 2). In agreement with their role in transcription, 13 out of the 83 cap-d3 DEG are transcription factors (Table S3). We conclude that the influence of CAP-D3 directly on transcription is moderate. However, the DEG involvement in plant response to stress, and the high proportion of transcription factors indicate that CAP-D3 may influence transcription indirectly.

Arabidopsis CAP-D proteins are expressed in meristematic tissues in a cell cycle-dependent manner
CAP-D2 and CAP-D3 are highly expressed in meristems and cell cycle active tissues (flower buds, roots), but weaker in non-cycling tissues (mature leaves). Similarly, the condensin subunit genes CAP-H and SMC2 are highly expressed in dividing tissues (Fujimoto et al., 2005;Liu et al., 2002;Siddiqui et al., 2003). Sequences of 391 bp or 474 bp upstream of the start of CAP-D2 or CAP-D3, respectively, are sufficient to act as promoters. Longer fragments (>1000 bp) do not improve the expression of the reporter gene. Interestingly, the CAP-D2 and CAP-D3 promoters regions contain two previously predicted E2F binding sites (Schubert et al., 2013). E2F is a transcriptional activator of genes important for cell cycle progression. Together with retinoblastoma-related protein (RBR) and a dimerization partner, they control the transition from G1 to S phase. E2F sites are also present in the A. thaliana SMC2 promoter (Siddiqui et al., 2003). In mouse, CNAP1 (CAP-D2) is also a target of E2F (Verlinden et al., 2005). Considering the expression patterns, the promoter features and the comparison with other organism, it is plausible that the transcription of A. thaliana CAP-D2 and CAP-D3 is cell cycle-regulated. Introns, when affecting the expression of a gene, often enhance its expression by increasing the transcript amount or by inducing the expression in specific tissues (Rose et al., 2008;Parra et al., 2011;Heckmann et al., 2011). Nonetheless, the second intron of CAP-D2 could have intragenic regulatory sequences repressing the expression in non-dividing tissues. This is supported herein by the loss of GUS reporter expression in leaves of the Pro6 and Pro9 transgenic plants compared with Pro5, Pro7 and Pro8 plants, which do not carry the second intron. Moreover, our quantitative RT-PCR results showed low transcription of CAP-D2 in leaves. The second intron of the AGAMOUS gene is also responsible to inhibit expression in vegetative tissues, and drives its correct expression in flowers (Sieburth and Meyerowitz, 1997).

The subunit composition of Arabidopsis condensin I and II is similar as in other eukaryotes
Protein immunoprecipitation (IP) from flower bud extracts confirmed already the presence of the subunits for condensin I and condensin II in A. thaliana (Smith et al., 2014). Nonetheless, these IPs were performed with anti-SMC4, which would target both condensin complexes, and therefore could not determine the exact composition of condensin I and II. Our data based on affinity purification combined with mass spectrometry support that in A. thaliana both condensin complexes are present, and that their subunit composition is identical to those of other organisms (Hirano, 2012a). Notably, A.thaliana is the only species in which two SMC2 homologs have been predicted and described (Cobbe and Heck, 2004;Siddiqui et al., 2003). Both, SMC2A and SMC2B can be active, but SMC2A accounts for most of the SMC2 transcript pool (Siddiqui et al., 2003). Both SMC2A and SMC2B interact with the other condensin subunits in vegetative and somatic tissues (Smith et al., 2014;this study). In human cells and Drosophila, CAP-D3 interacts with RBR and promotes the correct chromosomal localization of condensin II (Longworth et al., 2008). In A. thaliana, this interaction is likely not conserved, since we could not detect RBR among the proteins that co-purified with CAP-D3. In human, Cdc20, a component of the anaphase-promoting complex E3 ubiquitin ligase, interacts and regulates CAP-H2 (Kagami et al., 2017). In Drosophila, CAP-H2 also interacts and is regulated by the Skp cullin-F-box SCF Slimb (Buster et al., 2013), an E3 ubiquitin ligase regulated by CSN (Hotton and Callis, 2008). In our affinity purifications, we also detected components of the ubiquitin-26S proteasome pathway. CULLIN 1 co-purified with CAP-D2 and CSN1 with CAP-D3 in all replicates. CSN3 and CSN4 also co-purified with CAP-D3 in the three triplicates but also in 3 out of 115 of the non-specific proteins affinity purifications (data not shown). CULLIN1 and CULLIN3 were present in two of the CAP-D3 triplicates. These data suggest that in A. thaliana, ubiquitination could be involved in the regulation of the condensins. A screen for functional partners of condensin in yeast identified, among others, two chromatin remodeling proteins and a histone deacetylase, as collaborators of condensin for chromosome condensation (Robellet et al., 2014). In line with that, we identified chromatin remodeling enzymes (CHR17, CRH19 and BRAHMA), histone chaperones (NAP1;1 and NAP1;2), a histone deacetylase (HDC1) and a histone acetyltransferase (ELO3) in the affinity purification experiments with CAP-D2 and CAP-D3. All of them are chromatin modifiers important for plant development (Perrella et al., 2013;Skylar et al., 2013;Gentry and Hennig, 2014).

Condensin I is located within the cytoplasm and nuclei during interphase
During interphase, in vertebrates the most commonly described localization of condensin I is exclusively in the cytoplasm (Hirota et al., 2004;Ono et al., 2004;Gerlich et al., 2006;Hirano, 2012a). However, some reports regarding Drosophila, chicken and human cell cultures described the localization of condensin I additionally within the nucleus (Schmiesing et al., 2000;Savvidou et al., 2005;Zhang et al., 2016). In A. thaliana protoplasts and N. benthamiana epidermal leaves, we observed CAP-D2, CAP-G and CAP-H EYFP-fusion proteins in the cytoplasm as well as the nucleus. The cytoplasmic localization of CAP-H was already described (Fujimoto et al., 2005), but not yet its nuclear localization. In stable A. thaliana transformants carrying CAP-D2 or CAP-D3 tagged at its N-or C-terminus to EYFP and in stable cap-d3 mutants carrying CAP-D3-EYFP, the fusion proteins could not be visualized, neither directly nor indirectly. The constructs are functionally active, since they work in A. thaliana and in N. benthamiana after transient transformations. In addition, the CAP-D3-EYFP construct was able to complement the centromeric phenotype of the cap-d3 SALK mutants. Similar problems have been described for GFP-PATRONUS1 A. thaliana transformants (Zamariola et al., 2014). These authors suggested that the reason behind could be the low expression or stability of the PATRONUS protein due to the presence of an APC/C degradation box. However, in CAP-D2 no APC/C degradation box exists. The detection of CAP-D2 in leaves from A. thaliana wild-type plants by Western blot was also not possible. This may be due to a low protein level in wild-type leaves since the transcript level in leaves is very low. By Western blot the CAP-D2 protein was detectable in protoplasts only when constitutively overexpressed. Similarly, in Drosophila the detection of condensin from extracts of non-dividing tissues was also not possible (Cobbe et al., 2006).

CAP-D3 may influence interphase chromatin arrangement and transcription, but not histone modifications
In Drosophila, CAP-D3 and CAP-H2 are needed to form compact chromosomes (Hartl et al., 2008;Bauer et al., 2012). Condensins via maintaining chromatin condensation may also maintain nuclear shape and size, as indicated after SMC2, CAP-H2 and CAP-D3 depletion in human cells (George et al., 2014). In embryonic stem cells of mice, the depletion of SMC2 causes chromatin decondensation as well as the increase of the nuclear volume (Fazzio and Panning, 2010). On the other hand, in C. elegans, the depletion of SMC4, CAP-G2 or HCP-6 (CAP-D3) does not change the chromosome territory volumes (Lau et al., 2014). In A. thaliana, previous studies based on FISH suggested an influence of CAP-D3 on the formation of the top arm 1 interphase chromosome territories and sister chromatid cohesion (Schubert et al., 2013). Using FISH probes against a smaller part of chromosome 1 bottom arm, we could not detect an increase of the hybridization signal area in the cap-d3 mutants compared to wild-type plants. The differences could be explained by labeling only one fourth of the chromosome arm by FISH, while in the previous study the whole chromosome arm (without pericentromeric heterochromatin) was visualized. The different methods used to quantify the dispersion of the interphase chromatin could be another reason. The degree of chromatin condensation within nuclei may depend on the type of tissue (Tessadori et al., 2007;van Zanten et al., 2011). Light (Bourbousse et al., 2015, drought, temperature, and salinity stress, as well as toxic components, energy-rich radiation and chemically induced DNA damage may also induce dynamic structural changes in plant chromatin (reviewed in Probst & Mittelsten-Scheid, 2015). Even compressive stress has the potential to reorganize chromatin (Damodaran et al., 2018;Xia et al., 2018). Thus, these factors have also the potential to influence chromatin condensation in the CAP-D mutants. Although we could not confirm the euchromatin dispersion in the cap-d3 mutants, it cannot be excluded that CAP-D3 is involved in the organization of chromosome territories as found in Drosophila (Bauer et al., 2012;Hirano, 2012b). The mutants used in our analysis (cap-d3 SAIL and cap-d3 SALK) have knockdown alleles, meaning that there is still a truncated transcript that could produce a partially functional CAP-D3 protein (Schubert et al., 2013). In addition to its role in chromosome compaction, condensin II has been described to influence transcription (Longworth et al., 2012;Dowen et al., 2013;Yuen et al., 2017). Although the A. thaliana cap-d3 mutants showed only moderate transcriptional changes, CAP-D3 might still affect the expression of genes involved in transcription and response to stress. This conclusion arises from our observation on the cap-d3 mutants which die sooner than wild-type plants after stress, such as pathogen infection. Interestingly, gross chromosome rearrangements altering the genome topology do not alter gene expression in Drosophila (Ghavi-Helm et al., 2018). Even a budding yeast strain, after merging all 16 chromosomes into a single one, revealed a nearly identical transcriptome and similar phenome profiles as wild-type strains (Shao et al., 2018). Thus, chromatin structure changes as induced in the A. thaliana cap-d3 mutants seem to influence the global transcription only slightly. Wang et al. (2017) showed that A. thaliana SMC4, but not CAP-D3, is important to maintain the repression of pericentromeric retrotransposons independent of DNA methylation. Accordingly, we observed no increased retrotransposon transcription in any of the cap-d3 mutants. Moreover, in accordance with our observations for both cap-d3 mutants, the protein-coding genes up-regulated in smc4 mutants are mainly involved in flower development, reproductive processes and DNA repair, and are distributed all over the genome (Wang et al., 2017). However, we observed in the cap-d3 mutants a differential expression of genes involved in transcription and stress response. This difference could be due to the combined effects of both condensin complexes I and II in the smc4 mutants, while in our cap-d3 mutants only condensin II is compromised. Posttranslational histone modifications may affect the structure and stiffness of interphase nuclei, and decondensed euchromatin correlates with less rigid nuclei (Chalut et al., 2012;Krause et al., 2013;Haase et al., 2016). In human HeLa cells histone methylation, but not acetylation, contributes to the stiffness and structure of condensed mitotic chromosomes (Biggs et al., 2019). Histone acetylation relaxes chromatin allowing different protein complexes to access DNA. Thus, acetylation is associated with transcription, and hypoacetylation with transcriptional repression (Wang et al., 2014). It seems that the unaltered degree and pattern of histone acetylation reflects an only moderate effect on transcription as we observed in the capd-3 mutants.

CAP-D proteins prevent heterochromatin clustering
CAP-H2 promotes the spatial separation of heterochromatic regions in Drosophila during interphase (Bauer et al., 2012;Buster et al., 2013). Correspondingly, in A. thaliana, depletion of CAP-D3 results in centromere clustering at interphase (Schubert et al., 2013). We confirmed this interphase phenotype and found that CAP-D3 depletion also results in the clustering of the 45S rDNA loci but not of the 5S rDNA sites. A differential behavior of rDNA was also found in protoplasts of A. thaliana. 45S rDNA remains condensed while the 5S rDNA decondenses during protoplast formation (Tessadori et al., 2007). 5S and 45S rDNA are transcribed by RNA polymerases III and I, respectively (Layat et al., 2012). Therefore, the different clustering behaviors of both rDNAs in the cap-d3 mutants could be due to their different structural and functional properties. Moreover, condensin of fission yeast, which is similar to condensin I, binds to RNA polymerase III transcribed genes (5S rDNA and tRNA), and mediates their localization near the centromeres (Iwasaki et al., 2010). The nuclear and chromocenter phenotype which we observed in the cap-d3 mutants differs from previous reports (Moissiard et al., 2012;Sakamoto and Takagi, 2013;Tatout et al., 2014;Poulet et al., 2017;Wang et al., 2017). The chromocenters cluster and localize at the nuclear periphery, but do not decondense, the nuclear area does not change compared to that of wild-type, and the general degree of DNA and histone methylation is unaffected. Moreover, hypomethylation, linc and morc mutants do not show transcriptional silencing of centromeric and pericentromeric repeats, and of silenced genes (Moissiard et al., 2012;Poulet et al., 2017). In contrast, CAP-D3 has little effect on silencing, because no increased transcription of transposable elements was detected in cap-d3 mutants (Wang et al., 2017). MORC, CRWN and LINC proteins localize close to the chromocenters, MORC foci adjacent to the chromocenters (Moissiard et al., 2012), CRWN1 and CRWN4 at the nuclear periphery (Sakamoto and Takagi, 2013), and the LINC complex in the nuclear envelope (Tatout et al., 2014). Conversely, CAP-D3 influences the arrangement of the chromocenters but localizes exclusively in euchromatin during interphase (Schubert et al., 2013). Therefore, CAP-D3 has mainly a structural role during interphase and affects the clustering of chromocenters without localizing close to them. Statistical analysis detected a more regular, than a completely random spatial centromere and chromocenter distributions in animal and plant nuclei. This suggests that repulsive constraints or spatial inhomogeneities influence the 3D organization of heterochromatin (Andrey et al., 2010). Computer simulation modeling of A. thaliana chromosomes as polymers predicts that the position of the chromocenters in the nucleus is due to non-specific interactions (de Nooijer et al., 2009). The simulated chromosomes exhibit chromocenter clustering except for the so-called Rosette chromosomes, in which the euchromatin loops emanate from the chromocenter and thus prevent chromocenter clustering . Indeed, depletion-attraction forces predict that big particles in an environment crowded with small particles will tend to cluster together (Marenduzzo et al., 2006). This situation can be applied to the nucleus where the chromocenters act as big particles and euchromatin as small particles. If association is not prevented, the chromocenters will cluster. In cap-d3 mutants we observed chromocenter clustering but barely chromosome territory dispersion. During mitosis, CAP-D3 is needed to confer the rigidity of the chromosome arms (Green et al., 2012) and human condensin controls the elasticity of mitotic chromosomes (Sun et al., 2018). We suppose, that during interphase, CAP-D3 localizes in euchromatin, possibly along the euchromatic loops, mediating the rigidity which is needed to keep the chromocenters away from each other. In case of lacking or functionally impaired CAP-D3, the loops may be not stiff enough to prevent the chromocenter clustering while the chromosome territories may mainly keep their structures (Fig. 9). The finding that condensed chromatin resist to mechanical forces, whereas decondensed chromatin is more soft (Maeshima et al., 2018) supports the idea that the stiffness of chromatin is an important feature to organize cell nuclei. Our observation that the degree of methylation and acetylation is not altered in the cap-d3 mutants suggests that these post-translational histone modifications are not required for the rigidity of interphase chromosome territory structures.

Plant material and stable transformation
All Arabidopsis thaliana (L.) Heynh lines and control plants are in Columbia-0 (Col-0) background.
The T-DNA cap-d3 (SAIL_826_B06, SALK_094776) insertion lines were previously described and selected in our laboratory (Schubert et al., 2013). Seeds were sown in soil and germinated under shortday conditions (16 h dark/8 h light, 18-20°C) and then transferred to long-day conditions (16 h light/ 8 h dark, 18-20°C) before bolting. The lines were genotyped by PCR using the primers listed in Table  S4. The presence of the T-DNA was further confirmed by sequencing. A. thaliana stable transformants were generated by the floral dip method (Clough and Bent, 1998). For selection of primary transformants, the seeds were sterilized and plated on ½ Murashige and Skoog (MS) basal medium (Sigma) supplemented with the adequate antibiotics when required and grown in a growth chamber under-long day conditions.

Transcript quantification
For transcript quantification total RNA was extracted from leaves, roots, 7 days-old seedlings and flower buds with the RNeasy Plant Mini kit (Qiagen) following manufacturer's instructions. All RNA samples were treated with TURBO DNAse (Thermo Fisher Scientific) and tested for DNA contamination by PCR. Reverse transcription was performed using 250 ng of total RNA and the RevertAid H Minus First Strand cDNA Synthesis kit (Thermo Fischer Scientific), with oligo(dT)18 primers, according to manufacturer's instructions. The quality of the cDNA was checked with a PCR targeting EF1B mRNA (Elongation factor 1 ). Quantitative RT-PCRs for CAP-D2 and CAP-D3 transcripts were done in triplicates and from three independent biological samples using SYBR™ Green PCR Master Mix (Thermo Fischer Scientific) in a 7900HT Fast Real-Time PCR System (Applied Biosystems). For each reaction, 0.5 µl of cDNA template and 0.6 mM primers (Table S4) were used in 10 µl. PPA2 and At4g26410 (Kudo et al., 2016) were used as reference genes for data normalization and the data were analyzed with the Double Delta Ct method (Livak and Schmittgen, 2001).

CAP-D2
and CAP-D3 promoter::GUS reporter lines and β-glucuronidase activity assay Different lengths of the promoter regions of both CAP-D2 and CAP-D3 were cloned between the SalI and NotI restriction sites of the pEntr 1A plasmid (Invitrogen). The sequences were amplified from gDNA with the primer pairs D2-1156F/D2ProR for the Pro4 fragment, D2-1156F/D2Int1R for Pro5, D2-1156F/D2Int2R for Pro6, D2-392F/D2ProR for Pro7, D2-392F/D2Int1R for Pro8, D2-392F/D2Int2R, for Pro9, D3-1318F/D3ProR for Pro10 and D3-474F/D3ProR for Pro11 (Table S4). The fragments were subcloned upstream of the GUS reporter gene in the pGWB633 plasmid (Nakamura et al., 2010) using the Gateway LR Clonase II enzyme mix (Invitrogen) following manufacturer instructions. Constructs were transformed into A. thaliana and stable transformants were selected in ½ MS (Sigma) with 16 mg/L PPT (Duchefa). One month after sowing, the plantlets were stained for GUS activity according to Jefferson et al. (1987) with small modifications. Plantlets were collected in 15 ml tubes containing 1% X-Glu (5-Bromo-4-Chloro-3-indolyl--D-Glucopyranoside; Duchefa) in 0.1 M phosphate buffer (pH 7.0). To facilitate the penetration of the solution in the material, the tubes with the plant material and the staining solution were exposed to vacuum for 5 min and incubated overnight at 37°C. Next day, the staining solution was replaced by 70 % ethanol and incubated 20 min at 60°C. This step was repeated until the chlorophyll was removed. The stained material was preserved in 70% ethanol at 4C° and analyzed under a stereo microscope.

Condensin subunit EYFP-fusion constructs
The 3942 bp and 4245 bp long cDNA sequences of CAP-D3 and CAP-D2 respectively, were synthesized and cloned into pEntr 1A (Invitrogen) by a DNA-Cloning-Service (Hamburg, Germany). The 3153 bp and the 2013 bp long cDNA sequences of CAP-H and CAP-G were amplified from flower bud cDNA with the primer pairs CAPH_pentry_f/CAPH_pentry_r and CAPG_pEnt_f/CAPG_pEntr_r (Table S4), respectively, and cloned between the SalI and Notl sites of the pEntr 1A plasmid (Invitrogen). Once in the pEntr 1A plasmid, the coding sequences of the genes of interest were subcloned into pGWB641 and pGWB642 plasmids (Nakamura et al., 2010) using Gateway cloning (Invitrogen). The generated expression cassettes contained the proteins of interest fused to EYFP C-terminally for the pGWB641 constructs (CAP-D2_EYFPc, CAP-D3_EYFPc, CAP-G_EYFPc and CAP-H_EYFPc) or N-terminally for the pGWB642 constructs (CAP-D2_EYFPn and CAP-D3_EYFPn), and both were under the control of the cauliflower mosaic virus 35S promoter. As a control (Control_EYFPc), a plasmid containing only EYFP under the control of the 35S promoter was generated.

Condensin I subunit localization in A. thaliana protoplasts and N. benthamiana
Isolation and transformation of A. thaliana leaf protoplasts were performed as described in Yoo et al., (2007). To improve the visualization of the fusion proteins, the transformed protoplasts were fixed in 4% formaldehyde in 1×PBS, washed in 1×PBS and centrifuged at 400 rpm for 5 min (Shandon CytoSpin3, GMI inc) onto a microscopic slide. The slide was directly used for immunostaining against EYFP. Nicotiana benthamiana leaf cells were transformed as described in Sparkes et al., (2006). Agrobacteria carrying the constructs of interest were grown in YEB medium with suitable antibiotics to an OD 600 of 1 and resuspended in infiltration medium (10 mM MES, 10 mM MgCl 2 , pH 5.6, 3.3 mM acetosyringone). N. benthamiana leaves of 2 to 3 weeks old plants were infiltrated with the Agrobacterium suspension using a syringe without needle and analyzed 2 to 4 days later.

Antibody production
The sequence between 2743 and 4248 bp of CAP-D2 (501 C-terminal amino acids) was amplified from A. thaliana flower bud cDNA with the D2CtSalI_F and D2CtNotlI_R primers (Table S4). The fragment was cloned between the SalI and NotI restriction sites of the pET23a(+) plasmid (Novagen) resulting in a pEt23_CAP-D2_Ct construct which contains the cassette T7 promoter::T7 tag-CAP-D2Ct-His tag. The construct was transformed into E. coli BL21 cells and the expression of the transgene induced with 1 mM IPTG (isopropyl--D-thiogalatopyranoside, Sigma-Aldrich). The recombinant protein was purified with agarose beads that bind specifically to the His-tag (Ni-NTA Agarose, Qiagen) following the purification hybrid method from the ProBond purification system (Thermo Fisher Scientific). The purified recombinant protein was used to produce anti-CAP-D2 polyclonal antibodies in rabbit (Udo Conrad, Phytoantibody group, IPK, Gatersleben, Germany). Two rabbits were immunized with the recombinant proteins and the anti-serum collected after four immunizations. The anti-CAP-D2 serum was used directly for Western blot.

Affinity purification and analysis of GS-tagged CAP-D2 and CAP-D3 from PSB-D cells
The cDNA sequences of CAP-D3 and CAP-D2 were synthesized and cloned into pCambia 2300 35S GS-Ct by the DNA-Cloning-Service (Hamburg, Germany) resulting in the constructs pCambia2300_CAP-D2_GS and pCambia2300_CAP-D3_GS. The A. thaliana ecotype 'Landsberg erecta' cell suspension (PSB-D) was transformed as described (Van Leene et al., 2011). CAP-D2-GS and CAP-D3-GS were affinity purified following the protocol described (Dürr et al., (2014). For mass spectrometry, the eluted proteins were separated in a 10 % polyacrylamide gel and digested with trypsin. Mass spectrometry and data analysis were performed according to Antosz et al., (2017). Protein Scape 3.1.3 (Bruker Daltonics) in connection with Mascot 2.5.1 (Matrix Science) facilitated database searching of the NCBInr database. Three independent affinity purifications were performed. A MASCOT score of minimum 100 and the presence in at least two of the purifications were considered as criteria for reliable protein identification. The experimental background (contaminating proteins that co-purified with the unfused GS-tag) and non-specific interactions (proteins that co-purified independently of the bait used) were subtracted. The list of non-specific A. thaliana proteins is based on 543 affinity purification experiments using 115 different baits (Van Leene et al., 2014).

Nuclei preparations
A. thaliana nuclei from differentiated leaf cells were isolated and flow-sorted according to their ploidy level as described (Weisshart et al., 2016) in a BD INFLUX Cell Sorter (BD Bioscience). The nuclei were sorted based on their DNA content in 2C, 4C, 8C and 16C ploidy fractions. Twelve µl of 4C sorted nuclei and the same amount of sucrose buffer (10 mM Tris, 50 mM KCl, 2 mM MgCl-6H 2 O, 5% sucrose, pH 8.0) were placed on a slide. The slides were directly used or stored at -20°C. A. thaliana nuclei were embedded in acrylamide to preserve their 3D structure following the procedure described by Kikuchi et al., (2005) with modifications. Twelve µl of nuclei suspension were mixed on a slide with 6 µl of active 15% acrylamide embedding medium (15% acrylamide/bisacrylamide (29:1), 15 mM PIPES, 80 mM KCl, 20 mM NaCl, 2 mM EDTA, 0.5 mM EGTA, 0.5 mM spermidine, 0.2 mM spermine, 1 mM DTT, 0.32 M sorbitol, 2% APS and 2% Na 2 SO 3 ). A coverslip was carefully placed on top of the acrylamide-nuclei mixture and let to polymerize 30 min at room temperature. The coverslip was then removed letting a thin pad of nuclei embedded in acrylamide on the slide that was directly used for FISH.

Preparation of squashed A. thaliana roots
A. thaliana seedlings were fixed in 4% paraformaldehyde in phosphate-buffered saline buffer (1×PBS buffer). The seedlings were washed in 1×PBS buffer and digested for 30 min at 37°C in an enzyme mix (0.5% pectolyase (Sigma), 0.5% cytohelicase (Sigma), 0.35% cellulase (Calbiochem), 0.35% cellulase (Duchefa) in 1×PBS buffer. After removal of the enzyme solution and washing in 1×PBS, the root tips were transferred to a slide and squashed between coverslip and slide. The liquid nitrogen frozen coverslip was lifted and the slide directly used for immunostaining.

Probe preparation and fluorescence in situ hybridization (FISH)
The probes were generated by: (i) PCR for the 180 bp centromeric repeat (pAL; Martinez-Zapater et al., 1986), (ii) from a plasmid for the 5S rDNA probe (pCT4.2;Campell et al., 1992), (iii) from BACs containing the 45S rDNA repeats (BAC T15P10), and (iv) for painting a part of chromosome territory 1 bottom (CT1B) from BACs arranged in contigs (BACs F11P17 to F12B7). The BACs were obtained from the Arabidopsis Biological Resource Center (Ohio, USA). The probes were labeled with modified dUTPs conjugated with Texas-red (Invitrogen) or Alexa-488 (Invitrogen) by nicktranslation. The FISH was performed as previously described (Pecinka et al., 2004).

Indirect immunofluorescence labeling
Nuclei and chromosome preparations were washed in 1×PBS and incubated for 30 min at 37°C in a moist chamber with 30 µl blocking buffer (4% BSA, 0.1% Tween-20 in 1×PBS) to reduce nonspecific antibody binding. After three washes in 1×PBS, the slides were incubated with the primary antibodies diluted in antibody buffer (1% BSA, 0.1% Tween-20 in 1×PBS) overnight at 4°C. Next day, the slides were washed in 1×PBS again and incubated with the secondary antibodies in antibody buffer for 1 h at 37°C. After incubation, the preparations were washed in 1×PBS, dehydrated in an ethanol series (70%, 90% and 96% ethanol for 2 min each) and counterstained with DAPI in antifade (Vectashield). All primary and secondary antibodies, and the dilutions used are listed in Table S5. Immunolocalization of 5-methyl-cytosine requires an initial DNA denaturation of the specimen. Therefore, slides with sorted nuclei were denatured in 70% formamid in 2×SSC for 2 min at 70°C. The preparations were dehydrated in ice cold 70% and 96% ethanol for 5 min each, and air-dried. Subsequent blocking and antibody incubation were carried as described above.

Microscopy and image analysis
Wide-field fluorescence microscopy was used to evaluate and image the nuclei and chromosome preparations with an Olympus BX61 microscope (Olympus) and an ORCA-ER CCD camera (Hamamatsu). When higher resolution was needed to analyze the spatial arrangement of the chromocenters, a super-resolution fluorescence microscope Elyra PS.1 and the software ZEN (Carl Zeiss GmbH) was used. Processing and analysis of microscopic image stacks were done with ZEN, Adobe Photoshop CS5 (Adobe) and Imaris 8.0 (Bitplane) software. The CT1B signals were quantified on 16-bit gray scale microscopic images using ImageJ v1.50i (Schneider et al., 2012). The images were taken from preparations of flow-sorted nuclei. Since this technique flattens the nuclei, they were considered as two-dimensional. Within each dataset all images were treated the same way after using the same acquisition parameters. For the CT1B signal image dataset, the background was subtracted with the option 'Rolling ball' set at 25 pixels and the delimitation of the region of interest (ROI) with the RenyiEntropy threshold. For the nuclear area image dataset (measured based on DAPI staining), the background was not subtracted, and the nuclear area was delimited as a ROI also with the RenyiEntropy threshold. The area of each ROI was automatically measured by the program.

Southern blot analysis
Five µg of genomic DNA from A. thaliana leaves was digested with either HpaII or MspI (Thermo Fischer Scientific). The DNA was gel-separated and transferred onto a nylon membrane (Hybond XL, Amersham). The 32 P-labelled centromeric 180 bp repeat pAL was used for Southern hybridization and the signals were detected by autoradiography. The A. thaliana centromeric pAL probe was generated by PCR and 32 P-labeled according to manufacturer´s instructions (Deca-Label DNA labelling Kit, Thermo Scientific).

cap-d3 RNA-seq and in silico transcriptome analysis
cap-d3 SAIL, cap-d3 SALK and control (Col-0) seeds were sown in soil and grown under short day conditions. RNA was extracted with the RNeasy Plant Mini Kit (Qiagen) from 50 mg of pooled 4 weeks old plantlets cut above the root. For each of the three A. thaliana genotypes five independent RNA extractions were performed and the RNA integrity of the samples was measured in a 2100 Bioanalyzer (Agilent). The four RNA samples of each genotype with the highest RIN (RNA Integrity Number) were used for library preparation and RNA sequencing (NGS platform, IPK Gatersleben, Germany). The libraries were prepared with a TruSeq RNA Library Kit (Illumina) unstranded and sequenced in a HiSeq2000 system (single 100 bp reads). The quality of the RNA-seq reads were assessed with FastQC v0.11.4 (Babraham Bioinformatics) and adaptors trimmed with Trimmomatic v0.32 (Bolger et al., 2014). After a second quality check in FastQC, the reads were aligned with GSNAP v2016-05-25 (Wu and Nacu, 2010) against the Arabidopsis TAIR10 genome and the gene counts calculated with HTseq v0.6.1 (Anders et al., 2015). Differential expression analyses were performed using the DESeq2 1.14.0 Bioconductor package (Love et al., 2014). Genes were considered differentially expressed (DEG) when they had a Benjamini-Hochberg-adjusted-P value ≤ 0.05 and a log 2 -fold change ≤ -1 or ≥ 1. These steps were performed through Galaxy (Afgan et al., 2018).Genes detected as differentially expressed for both cap-d3 mutants were considered as the genes associated to CAP-D3 defective proteins independently of the specific mutation. Gene enrichment was analyzed with agriGO v1.2 (Du et al., 2010). The analysis of the transcription factors present in cap-d3 DEG was perform with the Arabidopsis Transcription Factor Database (AtTFDB) from the Arabidopsis Gene Regulatory Information Server (AGRIS; Yilmaz et al., 2011).

Gene and protein identification numbers
Sequence data from this study can be found in The Arabidopsis Information Resource (TAIR, www.arabidopsis.org) or National Center for Biotechnology Information (NCBI, www.ncbi.nlm.nih.gov/) databases under the following gene identification numbers: CAP-D2, AT3G57060; CAP-D3, AT4G15890; CAP-G, AT5G37630; CAP-H, AT2G32590.          (Marenduzzo et al., 2006). Consequently, the chromocenters cluster (right). Figure S1. In silico analysis of A. thaliana CAP-D2 and CAP-D3 expression. The results obtained with the Arabidopsis eFP Browser 2.0 (bar.utoronto.ca) revealed a similar expression level for both genes with high (red), medium (orange) and low (yellow) expression in different organs and developmental stages.

Figure S2. Protein-protein interaction network of CAP-D2 (condensin I) and CAP-D3 (condensin II). Both
A. thaliana CAP-D2 (a, c) and CAP-D3 (b, c) proteins (red) interact potentially with the other coiled-coil condensin SMC complex components (green) and the condensin I-and condensin II-specific subunits (yellow). The network was generated by the STRING program (http://string-db.org/) analysis at scores >0.90 (a, b) and >0.70 (c), respectively. The black lines in between the proteins indicate the supporting evidence from experimental data available from different species. The dashed lines embrace the condensin I and II subunits in (c).

Figure S8. Immunolocalization of histone modifications in cap-d3 mutants and wild-type plants.
No differences were detected in 4C nuclei of wild-type (Wt) and the cap-d3 SAIL, cap-d3 SALK mutants tested with antibodies against histone H3K27me3 (euchromatic); H3K9me2 (heterochromatic); H3K9ac and with antibodies recognizing H3K14+18+23+27ac.          . Figure 9. Model explaining the function of Arabidopsis CAP-D3 in interphase nuclei. Two chromosomes are represented in blue and in red with euchromatin emanating loops from their pericentromeric chromocenters (rosette chromosome model; Fransz et al., 2002;de Nooijer et al., 2009). CAP-D3 (green circles) localizes along euchromatin creating the chromatin loops rigid to keep the chromocenters separated (left). In absence of CAP-D3 the chromatin loops are not stiff enough to counterbalance the depletion-attraction forces (Marenduzzo et al., 2006). Consequently, the chromocenters cluster (right).  Figure S2. Protein-protein interaction network of CAP-D2 (condensin I) and CAP-D3 (condensin II). Both A. thaliana CAP-D2 (a, c) and CAP-D3 (b, c) proteins (red) interact potentially with the other coiled-coil condensin SMC complex components (green) and the condensin I-and condensin IIspecific subunits (yellow). The network was generated by a STRING program (http://string-db.org/) analysis at scores >0.90 (a, b) and >0.70 (c), respectively. The black lines in between the proteins indicate the supporting evidence from experimental data available from different species. The dashed lines embrace the condensin I and II subunits in (c).      Figure S8. Immunolocalization of histone modifications in cap-d3 mutants and wild-type plants. No differences were detected in 4C nuclei of wild-type (Wt) and the cap-d3 SAIL, cap-d3 SALK mutants tested with antibodies against histone H3K27me3 (euchromatic); H3K9me2 (heterochromatic); H3K9ac and with antibodies recognizing H3K14+18+23+27ac.