Involvement of condensin in cellular senescence through gene regulation and compartmental reorganization

Senescence is induced by various stimuli such as oncogene expression and telomere shortening, referred to as oncogene-induced senescence (OIS) and replicative senescence (RS), respectively, and accompanied by global transcriptional alterations and 3D genome reorganization. Here, we demonstrate that the human condensin II complex participates in senescence via gene regulation and reorganization of euchromatic A and heterochromatic B compartments. Both OIS and RS are accompanied by A-to-B and B-to-A compartmental transitions, the latter of which occur more frequently and are undergone by 14% (430 Mb) of the human genome. Mechanistically, condensin is enriched in A compartments and implicated in B-to-A transitions. The full activation of senescence genes (SASP genes and p53 targets) requires condensin; its depletion impairs senescence markers. This study describes that condensin reinforces euchromatic A compartments and promotes B-to-A transitions, both of which are coupled to optimal expression of senescence genes, thereby allowing condensin to contribute to senescent processes.

OIS cells were further infected with lentivirus encoding one of the three shRNA constructs (#1, #2, and #3) against NCAPH2 or carrying an empty vector (control) and harvested 3 days after the lentivirus infection. P values were calculated by two-sided Student's t test, using biologically independent samples (n = 3, error bars represent the SD).
(b) Whole cell lysates of OIS cells with and without CAP-H2 KD were subjected to western blot analysis. CAP-H2 antibody (Bethyl Laboratories, A302-275A) was used for immunoblotting.
Tubulin serves as a loading control. Asterisk indicates non-specific band. Relative expression of CAP-H2 proteins after CAP-H2 KD (#1, #2, and #3) compared to expression without KD (control) was estimated.
(c) RT-qPCR to investigate NCAPH2 mRNA in growing (G) and OIS cells. P values were calculated by two-sided Student's t test, using biologically independent samples (n = 3, error bars represent the SD). (d) Distributions of CAP-H2, SMC1, CTCF, and Pol II binding sites at the indicated genetic elements, in growing cells. Numbers of total binding sites for the indicated proteins are shown at top. For the control, 10,000 loci were randomly selected from the entire genome and classified into the same categories.
(e) Comparison of CAP-H2, SMC1, CTCF, and Pol II binding sites between OIS and growing cells.
Colors represent the genetic elements annotated in panel d.
(f) Overlap of CAP-H2 binding peaks in OIS and growing cells.
(g) Overlap among binding sites of the indicated factors in OIS cells.
(h) Co-occupancy of genomic regions by the same factors as in panel g. Colors represent the genetic elements annotated in panel d.
(i) Changes in gene expression in OIS cells compared to growing cells, as determined by RNAseq data. Expression ratios between OIS and growing cells (Y axis) are plotted against average expression levels for respective genes (X axis). Red and blue dots indicate significantly up-(n = 842) and down-regulated (n = 865) genes, respectively (Methods).
(j) Average CAP-H2-FLAG and CAP-H2 binding patterns at the significantly up-(top, red, n = 808) and down-(bottom, blue, n = 842) regulated genes in OIS cells compared to growing cells; only genes > 2 kb were selected for analysis. Grey lines show average binding patterns at genes without significantly altered expression (n = 15,893 among 19,591 genes).
(k) Nascent transcripts visualized in OIS and growing cells. Cells were treated by an RNA polymerase II inhibitor, Triptolide (1 µM), for 1 hour and subsequently incubated with 5-ethynyl uridine (EU) for additional 1 hour. EU was incorporated into nascent RNA and reacted with Cy3azide via Click chemistry. This analysis was performed as detailed in Supplementary Methods, and distributions of nuclear nascent RNA signals are shown as boxplots (central bar represents the median with boxes indicating the upper and lower quartiles, and whiskers extend to the data points, which are no more than 1.5x the interquartile range from the box; two-sided Mann-Whitney U test). A.U., arbitrary unit.
(l) Effect of RNA polymerase inhibitor treatment (Triptolide and α-Amanitin) on nascent RNA levels determined by RT-qPCR. OIS cells were treated by the inhibitors and incubated with EU. EUincorporated nascent transcripts were reacted with biotin-azide via Click chemistry, purified using streptavidin beads, and subjected to RT-qPCR (Supplementary Methods). P values were calculated by two-sided Student's t test, using biologically independent samples (n = 3, error bars represent the SD).
(m) Effect of RNA polymerase inhibitor treatment on nascent RNA levels determined by RNA-seq.
OIS cells were treated with the RNA polymerase inhibitors, and nascent RNA prepared as in panel l were subjected to RNA-seq. Nascent RNA levels at CAP-H2 binding genes (n = 7,622) were determined as described in Supplementary Methods and plotted as boxplots (central bar represents the median with boxes indicating the upper and lower quartiles, and whiskers extend to the data points, which are no more than 1.5x the interquartile range from the box; two-sided Mann-Whitney U test).
(n) Effect of RNA polymerase inhibitor treatment on CAP-H2 binding. Inhibitor-treated OIS cells were subjected to CAP-H2 ChIP-seq analysis. Average CAP-H2 binding levels at CAP-H2 binding genes (n = 7,622) were calculated.
Nascent RNA levels and CAP-H2 enrichment at CAP-H2 binding genes in OIS cells treated by Triptolide (5 minutes) and α-Amanitin (1 hour) were normalized by those without the inhibitor treatment. Pearson's correlation coefficient (r) between relative nascent RNA levels and CAP-H2 ChIP-seq enrichment was indicated.

red) and growing (bottom left, blue) cells
Contact maps were generated as described in Methods. Histone H3K9me3 ChIP-seq data (GEO accession#, GSE38448[https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE38448]) in growing cells are shown at top. Note that global H3K9me3 patterns were similar between OIS and growing cells 2 . Difference of contact probabilities between OIS and growing cells are also shown as described in Fig. 2b. PCA scores in growing and OIS cells are shown to the right as described in Fig. 2c.

Supplementary Figure 4. SAHF formation and compartmental reorganization upon OIS
(a) Reproducibility of in situ Hi-C data. Correlation between the indicated contact maps was evaluated using the HiCRep program 3 . Contact maps at 500 kb resolution were compared, where genomic combinations within 100 Mb were included in the estimation. Bio indicates biological replicas. Contact maps from OIS and growing WI-38 cells 4 were generated using the same procedure as for the in situ Hi-C data. and whiskers extend to the data points, which are no more than 1.5x the interquartile range from the box; two-sided Mann-Whitney U test).
(a) ChIP-seq enrichment of the indicated proteins and epigenetic marks for the respective compartmental categories. Five hundred loci (40 kb bins) were randomly selected from each compartmental category and from the entire human genome (control). Average ChIP-seq enrichment was calculated for each specific category and control, and a ratio of average ChIPseq enrichment between each category and control was determined. The random sampling was repeated 100 times, and an average log2 ratio of ChIP-seq enrichment was calculated.
(b) Relative ChIP-seq enrichment between OIS and growing cells for the respective compartmental categories. Five hundred loci (40 kb bins) were randomly selected from each compartmental category. Average ChIP-seq enrichment in OIS and growing (G) cells was calculated for each category, and a ratio of the average ChIP-seq enrichment between OIS and G cells was determined. The same calculation was applied to the entire genome (control). These calculations were repeated 100 times, and an average log2 ratio of ChIP-seq enrichment from the control was subtracted from that from each compartmental category.  (h) Same as panel g, but for BJ growing and RS cells.
(i) Overlap of BA transitions between the IMR90 RS biological replicates (Bio1 and Bio2).
Genomic regions (40 kb bins) undergoing BA transitions upon RS were compared between the IMR90 RS biological replicates (hypergeometric distribution test). Genes (n = 318) located at genomic regions undergoing BA transitions that were common between the IMR90 RS biological replicates were subjected to GSEA (bottom) to obtain NES and P values.
(j) Same as panel i, but for the BJ RS replicates (Bio1 and Bio2).
(k) Overlap of AB transitions between the IMR90 RS biological replicas (Bio1 and Bio2). Genomic regions (40 kb bins) undergoing AB transitions upon RS were compared between the IMR90 RS biological replicates (hypergeometric distribution test).
(m) Overlap of AB transitions upon OIS and RS in IMR90 cells. Genomic regions undergoing AB transitions that were common between the IMR90 OIS biological replicates were compared to those shared between the IMR90 RS biological replicates (hypergeometric distribution test).
(n) Overlap of AB transitions upon RS in IMR90 and BJ cells. Genomic regions undergoing AB transitions that were common between the IMR90 RS biological replicates were compared to those shared between the BJ RS biological replicates (hypergeometric distribution test). Fig. 7a were subjected to IF experiments visualizing spindle microtubules.

Cells prepared as in
White arrows indicate cells with mitotic spindles. n = 212, 211 and 308 for growing; n = 247, 281 and 324 for Con; n = 254, 250 and 318 for shRNA #1; n = 255, 264 and 302 for shRNA #3. P values were calculated by two-sided Student's t test, using biologically independent samples (n = 3, error bars represent the SD).
(a) SASP genes significantly up-regulated in OIS cells compared to growing cells were ranked by expression ratios between CAP-H2 KD and OIS cells. OIS cells were subjected to CAP-H2 KD. Significantly up-and down-regulated genes in CAP-H2 KD cells compared to OIS cells were indicated by darker color. Genes in red color were subjected to RT-qPCR analysis (Fig. 8).
(b) Same as panel a, but p53 target genes were subjected to the analysis. (j) Numbers of TAD borders that were specific to OIS and CAP-H2 KD cells. TAD borders were predicted from the OIS (Bio1-3) and CAP-H2 KD (#1-#3) data. TAD borders conserved in at least two OIS replicates but not in CAP-H2 KD data were classified as OIS-specific borders; TAD borders conserved in at least two CAP-H2 KD data but not in OIS data were categorized as CAP-H2 KD-specific borders. Remaining TAD borders detected in at least two data sets were classified as non-specific borders.
(k) Significant overlap between CAP-H2 KD-specific borders and G-specific borders as predicted in Fig. 5a (hypergeometric distribution test).

Characterization of CAP-H2 antibodies
A condensin II-specific subunit, CAP-H2, was detected by CAP-H2 antibody and depleted using the three shRNA constructs (#1, #2, and #3; Supplementary Fig. 1a,b). NCAPH2 mRNA were more abundant in growing cells than OIS cells (Supplementary Fig. 1c). We tested two different CAP-H2 antibodies and found that CAP-H2 proteins were detected by the both antibodies (Bethyl Laboratories, A302-275A; Abgent, AP1973A) and were much decreased by CAP-H2 KD (Supplementary Fig. 1d). CAP-H2 detected by the Bethyl antibody was reduced after induction of OIS, whereas CAP-H2 detected by the Abgent antibody was enhanced, potentially suggesting that the Abgent antibody might recognize modifications of CAP-H2 proteins (Supplementary Fig.   1d). Furthermore, the Bethyl CAP-H2 antibody detected exogenous CAP-H2-FLAG proteins, again indicating that this antibody recognizes CAP-H2 proteins (Supplementary Fig. 1e). We also performed cellular fractionation analysis as previously described 1 and found that CAP-H2 interacting with chromatin was clearly detected by the Bethyl antibody (Supplementary Fig. 1f,g).
Therefore, we employed the Bethyl CAP-H2 antibody to map condensin II across the human genome.

Definition of condensin II binding sites in OIS cells
For OIS cells, binding peaks of endogenous CAP-H2 and exogenous CAP-H2-FLAG proteins were determined by ChIP-seq. ChIP-seq experiments were performed twice for the respective proteins (biological replicas): CAP-H2-FLAG (OIS#1), CAP-H2-FLAG (OIS#2), CAP-H2 (Bethyl 275 OIS#1), and CAP-H2 (Bethyl 275 OIS#2). CAP-H2-FLAG binding peaks conserved between two biological replicas were obtained. Subsequently, the CAP-H2-FLAG common peaks were compared to endogenous CAP-H2 peaks. If the CAP-H2-FLAG common peaks were overlapped with the endogenous CAP-H2 peaks in at least one of biological replicas, then they were defined as CAP-H2 binding sites (Supplementary Fig. 2c). For growing cells, binding peaks of endogenous CAP-H2 (Bethyl 275 G) and exogenous CAP-H2-FLAG (G) were determined by performing ChIP-seq experiments once for the respective proteins, and their common peaks were defined as CAP-H2 binding sites.

Classification of ChIP-seq binding sites
Transcriptional start sites (TSSs) of protein-coding genes were selected, and less than 1 kb regions upstream and downstream from TSSs were defined as TSS sections. Histone H3K4me3 -31 -peaks were previously mapped in OIS and growing cells 2 . TSS sections overlapped with H3K4me3 peaks were defined as active promoters, and others were classified as inactive promoters.
Non-coding RNA genes were defined as follows: Locations of tRNA genes were annotated at GtRNAdb (http://gtrnadb.ucsc.edu/genomes/eukaryota/Hsapi19/); From Ensembl database (https://useast.ensembl.org), lincRNAs, miRNAs, miscRNAs, snRNAs, snoRNAs and rRNAs were selected. Genomic regions encoding these genetic elements were defined as non-coding RNA genes. Fig. 1) Enhancers were defined based on histone H3K27ac peaks in OIS and growing cells 7 . H3K27ac peaks were determined by the parameters: FDRs < 0.01, P values < 0.0001, and fold enrichment > 4. All TSSs with H3K27ac peaks were eliminated, and remaining H3K27ac peaks were defined as potential enhancers. Super enhancers were defined using HOMER software using the option "-style super". Enhancers not overlapped with super enhancers were defined as typical enhancers.

Treatment of human cells with RNA polymerase inhibitors
RNA polymerase II inhibitors, Triptolide and α-Amanitin, were used to impair transcription in human cells, as described previously 8 . Final 1 µM Triptolide (Sigma Aldrich, T3652) or 5 µg/ml α-Amanitin (Sigma Aldrich, A2263) was added to culture medium, and cells were cultured for the indicated duration.

Calculation of LVS scores
LVS scores were calculated from ICE normalized contact matrixes at 40 kb resolution, as previously described (SVL scores) 9 . LVS scores were defined as follows:

(Sum of contact probabilities for genomic combinations within 2 Mb),
( where contact probabilities for combinations within 40 kb (= intra-bins along diagonal lines) were eliminated from the calculation.

Visualization of nascent RNA using 5-ethynyl uridine (EU)
Nascent RNA was monitored using EU as detailed previously 10  To quantify Cy3 signals in the nucleus, low magnified images were captured by Nikon 80i Upright Microscope with 20 x objective lens. Nuclear area was defined by DAPI signals, and total Cy3 signals were quantified for each nucleus using NIS-Elements Microscope Imaging Software (Nikon). As a negative control, cells were cultured without EU and subjected to the same procedure. Cy3 signals derived from cells not treated by EU were quantified and used as background. Nascent RNA was calculated as follows.
Total nascent RNA per nucleus = Total Cy3 signals in selected nuclear area -(Nuclear area x Average background signal density), where nuclear area was estimated by the number of pixels with DAPI staining. Average background signal density was calculated by total nuclear Cy3 signals divided by nuclear area (pixels) in negative control cells, as described above. Distributions of total nascent RNA per nucleus were plotted as boxplots for respective conditions and subjected to two-sided Mann-Whitney U test.

Co-visualization of nascent RNA and CAP-H2
Final 0.5 mM EU was added to medium and cultured for 1 hour. To co-visualize nascent RNA and CAP-H2 proteins, cells grown on coverslip were fixed with 1% pFA in 1 ml PBS containing 200 mM sucrose at room temperature for 10 minutes. After washing cells three times with 1 ml PBS, cells were permeabilized with 0.5% Triton-X100 in 1 ml PBS at 4ºC for 5 minutes and washed three times with 1 ml PBST. Cells were soaked in PBS containing 1% BSA at room temperature for 10 minutes and incubated with 1:1000-diluted mouse monoclonal anti-FLAG M2 (Sigma Aldrich) in PBS for 30 minutes. After washing cells three times with 1 ml PBST, cells were incubated with 1:1000-diluted Alexa Flour 488-conjugated anti-mouse IgG (Thermo Fisher Scientific) in PBS for 15 minutes. After washing cells three times with TBS buffer, cells on coverslip were subsequently subjected to Click reaction to visualize nascent RNA as described above. Genomic DNA was stained with 1 µM TO-PRO-3 (Thermo Fisher Scientific) in PBS at room temperature for 10 minutes. After washing cells twice with 1 ml PBST, the coverslip was mounted onto slide glass using ProLong Gold antifade mountant (Thermo Fisher Scientific).

Quantification of EU-incorporated nascent RNA by RT-qPCR and RNA-seq
Isolation of nascent RNA was performed as described previously 11 with slight modifications. Final 0.5 mM EU was added to medium and cultured for 1 hour. Total RNA was purified from EUtreated cells using RNeasy mini kit (Qiagen for 10 minutes, and twice with 500 µl TNE 0.2 buffer. RNA-coupled beads were suspended in 10 µl nuclease-free water and subjected to reverse transcription using High-Capacity cDNA Reverse Transcription kit (Thermo Fisher Scientific). cDNA was used as a template in subsequent qPCR with gene specific primers (Power SYBR Green master mix, Thermo Fisher scientific). In order to normalize cell numbers in respective samples, genomic DNA was quantified using total RNA as a template and p21 primers for qPCR (Supplementary Table 2).
To make nascent RNA-seq libraries, cDNA generated from streptavidin bead-bound RNA was subjected to second strand synthesis using NEBNext Ultra II Non-Directional RNA Second Strand Synthesis Module (New England Biolabs). Double-stranded DNA was purified using SPRI beads (Beckman Coulter), followed by library preparation using NEBNext Ultra II DNA Library Prep Kit for Illumina (New England Biolabs). Adaptor-ligated DNA fragments were amplified for 10 cycles using NEBNext Q5 Hot Start HiFi PCR master Mix and NEBNext multiplex oligos (New England Biolabs). Nascent RNA-seq libraries were sequenced on Illumina NextSeq 500 platform to obtain 75-bp single-end reads. Nascent RNA-seq data were obtained for OIS cells treated with either Triptolide for 5 minutes or α-Amanitin for 60 minutes and also for cells without RNA polymerase inhibitor treatment (control) and processed as described in the RNA-seq section.
Moreover, nascent RNA levels of CCL20, MMP1, IER3, and TRIB1 genes were quantified by RT-qPCR using the same nascent RNA samples. Please note that nascent RNA-seq data were strongly correlated with the RT-qPCR results: coefficient of determination R 2 = 0.984 for Triptolide (5 minutes), R 2 = 0.904 for α-Amanitin (60 minutes), and R 2 = 0.987 for no inhibitor treatment.