Breakdown of a key chromosome protein in stressed and aging mouse cells

We treated murine NIH/3T3 cells with a representative panel of genotoxic agents under conditions known to promote various types of DNA damage () as revealed by accumulation of phosphorylated histone variant H2A.X (γH2A.X) and stabilization of p53 (). We monitored the impact of various drug treatments on cell cycle by FACS (). Centromere architecture was assessed in single cells using immunofluorescence (IF) to follow CENP-A localization and DNA-FISH using probes specific for centromeric repeats termed minor satellites in the mouse. In untreated cells, CENP-A staining and minor satellite repeats adopted the typical punctate pattern in the vicinity of chromocenters, composed of pericentromeric major satellite repeats or visualized as dense DAPI staining (; top rows). We first focused on Etoposide (ETOP), a potent inducer of DNA double strand breaks (DSB), as a paradigm for studying the impact of DNA damage on centromeres. We found that CENP-A became remarkably mislocalized away from its normal location and occupied the periphery and inside of the nucleolus marked by B23/nucleophosmin signal (; bottom row). Likewise, ETOP treatment led to the disorganization of centromeric DNA, mainly characterized by stretched rather than punctiform minor satellite signals, while chromocenters remained globally less affected (; two bottom rows). Table S1 Figure S1A Figure S1B ²² Fig. 1A and B Fig. 1A Fig. 1B

To exclude an effect restricted to topoisomerase II inhibition by ETOP, we tested other genotoxic drugs that have various impacts on DNA () and on cell cycle (). We found that Zeocin (ZEO), Mitomycin-C (MMC) and, to a lesser extent, Hydroxyurea (HU) in the conditions used, led to CENP-A relocation away from centromeric foci (). In contrast, forced cell-cycle arrest in G2/M after Nocodazole (NOC) treatment, a poison of microtubules, was not sufficient to promote CENP-A delocalization (). Table S1 Figure S1B Figure S2A Figure S2A

Kinetics experiments following genotoxic stress showed that, while maintaining the typical punctate pattern nearby chromocenters during the first 4 hr of ETOP treatment, CENP-A became visibly mislocalized by 8 hr, and further occupied the nucleoplasm after 24 hr (). In contrast, other centromeric proteins like CENP-B and CENP-C mainly displayed the typical centromeric punctate pattern, despite being slightly enriched in the nucleoplasm (). Although CENP-A delocalization seemed to be a relatively early event in the response to DNA damage, we followed the appearance of cleaved Caspase-3, one of the earliest molecular events at the onset of apoptosis. We did not detect cleaved Caspase-3 before 24 hr of ETOP treatment (). The signal was also undetectable in cells treated for 4 hr that have recovered from stress after a 24 hr period (). Hence, CENP-A relocation does not seem to be a direct consequence of the activation of pro-apoptotic signals. Rather, it may precede cellular changes. Fig. 1C Figure S2B and C Figure S2D Figure S2D

Etoposide forms non-repairable DSBwhile cells progressively accumulate in G2/M (). Interestingly, whereas a 4 hr ETOP treatment did not promote visible CENP-A delocalization (), a 24 hr recovery period (4 + 24H) during which DNA damage signalling and γH2A.X signal persisted () and most cells arrested in G2/M () promoted CENP-A delocalization in more than 60% of the cells (; lower row). Similar data were obtained with 4 hr ZEO or MMC treatments followed by a 24 hr recovery period (), time point at which γH2A.X is still detectable (). These data support a link between persistence of DNA damage and centromeres loss of identity through CENP-A delocalization. To confirm this hypothesis, we treated cells with lower doses of ETOP (2 μM instead of 10 μM) to allow cells to repair DNA damage, as shown by γH2A.X that returned to baseline levels after a 24 hr recovery from the 4 hr ETOP treatment (). In addition, cells were only transiently arrested in the cell cycle when treated with lower doses of ETOP. Indeed, in contrast to treatment with 10 μM ETOP that led to sustained cell cycle arrest, treatment with 2 μM ETOP allowed cells to resume cell cycle after the recovery period as shown by the increased number of cells in G1 after a 24 hr recovery from the 4 hr treatment (). Under these conditions, the percentage of cells with delocalized CENP-A decreased from nearly 70% to less than 40%, suggesting that CENP-A reoccupies its normal location when DNA damage is repaired (), although we cannot exclude that undamaged cells with or normal CENP-A location take over in the culture during the recovery period. ²³ Figure S1B Fig. 1C Figure S2D Figure S1B Fig. 1C Figure S3A Figure S3B Figure S3B Figure S1B Figure S3A

To confirm the observations made by microscopy, we used Chromatin Immunoprecipitation (ChIP) that showed a decreased CENP-A occupancy at centromeric repeats after ETOP treatment relative to untreated control cells, with a 2.4 fold after 24 hr and almost 4 fold after 4 hr ETOP+ 24 hr recovery () that parallels the percentage of cells with delocalized CENP-A (). In contrast to CENP-A, centromere content in canonical histone H3 did not vary significantly following genotoxic stress (), whereas content in histone H4 showed a reproducible 1.5-fold reduction after 24 hr ETOP treatment. ChIP experiments also showed that CENP-A did not accumulate at ectopic sites in the nearby pericentromeric major satellite repeats, nor in ribosomal DNA repeats (rDNA) despite its accumulation close to nucleoli (). Fig. 1D Fig. 1C Fig. 1E Fig. 1D

The dynamics of parental nucleosomal and newly synthesized CENP-A is tightly regulated during cell cycle. We took advantage of the SNAP-tag technology that allows distinguishing the fate of parental or newly synthesized histones. We generated NIH/3T3 cell lines expressing the fused CENP-A-SNAP protein, and used quench-chase-pulse or pulse-chase imaging protocols (see the Methods section). Quenching of parental SNAP histones followed by a 24 hr chase of the non-fluorescent substrate in the presence or absence of ETOP, before labelling of newly synthesized CENP-A with the fluorescent substrate TMR-Star (), showed that ectopic CENP-A was correctly incorporated in centromeric chromatin of untreated cells (). In contrast, and consistent with cells arresting in G2 while CENP-A is incorporated in G1, newly synthesized CENP-A was not deposited in centromeres nor close to the nucleolus of ETOP-treated cells (). Fluorescent labelling of the SNAP-tagged CENP-A () before a 24 hr ETOP treatment showed that parental nucleosomal CENP-A histones were mostly relocated close to the nucleolus concomitantly with endogenous CENP-A, as shown by line scan analysis (, bottom panels). These experiments suggested that CENP-A dispersion in response to DNA damage results from eviction of parental nucleosomal histones. ^{24 25 26 27 28} Fig. 2A Fig. 2B Fig. 2B Fig. 2C Fig. 2D

CENP-A mislocalization was observed in genotoxic stress conditions. Thus, we assessed the contribution of ATM and ATR kinase signalling cascades that play central roles in cellular response to DNA damage. We found that inhibition of ATM (ATMi), but not that of ATR (ATRi), prevented stress-mediated CENP-A delocalization (). While DNA damage persists in the presence of genotoxic stressor, its signalling is impaired by ATMi. Interestingly, the reduced CENP-A delocalization observed with ATMi but not ATRi correlated with override of the G2/M cell-cycle block (). ¹⁹ Fig. 3A Figure S3C

We noticed the presence of a phosphorylation consensus site for ATM (SQ motif) at serine 30 of murine CENP-A. Mutation of this residue into a non-phosphorylatable alanine (S30A) did not prevent deposition of CENP-A-SNAP suggested by its presence at centromeric foci (). In contrast, delocalization of the parental CENP-A-SNAP S30A mutant version following ETOP treatment was clearly reduced compared to that of the CENP-A-SNAP WT protein (). Delocalization of total CENP-A upon DNA damage was equivalent between CENP-A-SNAP WT and S30A cell lines (50 and 60%, respectively) and comparable to that of endogenous CENP-A in untransfected cells (). However, within the population of cells with delocalized total CENP-A, the percentage of cells with delocalized SNAP-tagged CENP-A dropped from 70% for the WT to 20% for the S30A mutant (), suggesting that both the ATM signalling pathway and a phosphorylatable Ser30 are required for CENP-A eviction in response to DNA damage in association with cell-cycle arrest. Fig. 3B Fig. 3B Fig. 1C Fig. 3B

CENP-A delocalization was detectable only after 8 hr of genotoxic stress (). Therefore, we assessed the accumulation of centromeric transcripts, another feature of stressed cells reported in various stress conditions (Reviewed in refs,). Various genotoxic conditions led to a strong increase in levels of centromeric transcripts as measured by RT-qPCR, relative to untreated cells (). This is in contrast to non-genotoxic stressors like heat shock (HS) or ethanol (EtOH) that had only modest effects in conditions commonly used (), suggesting that increased transcription of centromeric repeats requires a context of DNA damage as evidenced by accumulation of γ-H2A.X and accumulation of the cyclin-dependent kinase inhibitor 1A p21Cip/CDKN1A (p21) (). Fig. 1C ^{29 30} Figure S4A Figure S4A Figure S4A Cip

Kinetic analysis in response to ETOP or ZEO revealed that levels of centromeric transcripts increased within a few hours, with a 5-fold increase by 2 hr of ETOP (; left panel) reaching levels up to a thousand-fold in 24 hr (; right panel). Tandemly repeated pericentromeric major satellite repeats were also induced although with a slower kinetics, with a 2 to 3-fold increase by 4 hr with either drug (; left panel), and reaching apparent lower levels compared to minor satellites transcription, with a 100-fold increase at 24 hr (; right panel). In contrast, telomeric or rDNA tandem repeats, interspersed repeats like long (LINE) or short interspersed nuclear elements (SINE) and transposable elements like intracisternal A-particle (IAP), did not show changes in their transcripts levels upon DNA damage (), suggesting the selective activation of centromeric repeats in response to genotoxic stress. Fig. 4A Fig. 4A Fig. 4A Fig. 4A Figure S4B

We performed RNA-FISH to assess accumulation of centromeric transcripts at the single cell level. In agreement with previous reports, centromeric transcripts were hardly detectable in control cycling cells (). In contrast, centromeric transcripts first accumulated into discrete dots typical of centromeric foci detected in the vicinity of chromocenters in almost half of the cell population after 8 hr of ETOP treatment, suggesting that they remain associated with their transcription sites. Spreading of the signal throughout the nucleoplasm appeared in 20% of the cells after 24 hr of treatment (), raising the interesting possibility that the impact of increased levels of centromeric transcripts may not be restricted to local perturbations and may affect other nuclear compartments and functions. ^{5 7} Fig. 4B Fig. 4B

Accumulation of murine minor satellites is regulated during cell cycle with a peak in G2/M phase. However, levels reached in G2-arrested ETOP- or ZEO-treated cells were always 10-fold higher than in NOC-arrested cells (). Moreover, high levels of centromeric transcripts were also observed in cells treated with HU and MMC () that accumulated in G1 or in S (). Hence, accumulation of centromeric transcripts is not a mere consequence of a cell-cycle arrest in G2 but rather the result of an increased transcription or stabilization of these transcripts in response to genotoxic insults. Indeed, inhibition of the transcriptional machinery using the RNA polymerase II (RNA Pol II) inhibitor Actinomycin D (ACTD) together with ETOP treatment abolished the observed strong accumulation of minor satellite transcripts () as well as CENP-A delocalization (), suggesting that genotoxic stress promotes transcriptional activation of centromeric repeats and that this activated transcription is required for CENP-A relocation. However, forced expression of minor satellite repeats carried out as beforeand leading, 24 hr post-transfection, to levels similar to that obtained with a 24 hr ETOP treatment, () was not sufficient to affect CENP-A endogenous localization (). ⁷ Figure S4A Figure S4A Figure S1B Figure S5A Figure S5B ⁵ Figure S5C Figure S5D

One of the main effects of DNA damage is the activation of ATM/ATR signalingand rapid stabilization of the transcription factor p53, which in turn activates downstream effectors for an appropriate cellular response to DNA damage. We found that neither ATMi nor ATRi affected levels of centromeric transcripts compared to ETOP treated cells (), suggesting that transcription of centromeric transcripts is activated in DNA damage conditions but independently of ATM/ATR signalling. It also confirmed that high levels of centromeric repeats are not sufficient to promote eviction of CENP-A since ATMi prevented CENP-A delocalization () but not accumulation of satellite repeat transcripts (). ^{17 20} Figure S6A Fig. 3A Figure S6A

We then subjected p53-null murine embryonic fibroblasts (MEF) to the two different doses of ETOP used previously, which led to increased levels of γH2A.X regardless the p53 context, indicating that ATM/ATR signalling is intact in these cells. We verified that the levels of p53 were undetectable (), and that transcription and protein levels of its downstream target p21were activated at both doses in NIH/3T3 cells with wildtype-p53 (WT-p53) but not in p53 null cells (p53) (). We found that the levels of minor satellite transcripts did not increase in the absence of p53 after 4 hr of ETOP treatment (and) and 14-times less compared to WT-p53 cells at later time points (), suggesting that minor satellite transcription was not delayed but rather severely impaired in a p53-null cellular context. Figure S6B Figure S6B Fig. 5A S6B Fig. 5A Cip −/−

In non-genotoxic conditions of p53 stabilization, using the small molecule inhibitor of MDM2/p53 interaction Nutlin-3a, transcription from minor satellite repeats was not enhanced, although a slight effect was observed on p53-direct target genetranscription (). In contrast, stabilization of p53 by Nutlin-3 potentiated ETOP-mediated transcriptional activation of minor satellite repeats by 3-fold (). ChIP assays showed that, in ETOP-treated cells, p53 had a substantial ability to bind to a consensus single binding site found in minor satellite repeats, although to a lesser extent than to the strong double binding sites described inpromoter (). This binding increased by 2-fold in cells treated by both ETOP and Nutlin-3 in correlation with the observed increased transcription (and). These results suggest that stabilized p53 may bind to non-canonical sites in minor satellite repeats, although this binding and the consequent activated transcription of centromeric repeats appears to require the DDR signalling. CDKN1A CDNK1A Figure S6C Figure S6C Fig. 5B Figs 5B S6C

CENP-A maintained its default localization in 80% of p53-null cells (), consistent with both increased transcription of minor satellite transcripts and genotoxic stress signalling being required for CENP-A delocalization. Interestingly, while maintaining the normal CENP-A localization, cells with a defective p53 checkpoint showed an increased micronuclei formation (, arrows; and) and accumulated in the culture with more than a 2n genomic content () indicative of defective mitosis and genomic instability. Fig. 5C Figs 5C Fig. 5D Figure S6D

Since activated centromeric transcription in stress conditions, but not centromeric transcripts themselves, was required for CENP-A delocalization, we further tested whether it could act through chromatin remodelling. We used siRNA-mediated knockdown of a panel of chromatin remodelers for which a number of studies have shown their role at centromeric and pericentromeric repeats (and references included). We also focused on the FACT complex, an ATP-independent histone chaperone first discovered as promoting transcriptional elongationthrough nucleosome destabilization. In addition, FACT was shown to co-purify with CENP-A nucleosomes, and its subunit SSRP1 to be required for centromeric localization of CENP-A. We found that the levels of centromeric transcripts in ETOP-treated cells were unaffected by reduced levels of the chromatin remodelers/chaperones tested (). In conditions in which we were able to obtain more than 50% reduction in transcripts levels () the percentage of cells with mislocalized CENP-A after ETOP treatment was largely decreased depending on the remodelling/chaperone factor tested (). The most striking result was obtained following knockdown of SSRP1 expression, confirmed by a strong decrease in the levels of both mRNA and protein (), which resulted in less than 10% of cells with delocalized CENP-A after a 24 hr ETOP treatment compared to almost 50% in cells transfected with control siRNAs (). As a whole, these data revealed cooperation between activated transcription and chromatin chaperones/remodelers to relocate CENP-A in DNA damage conditions. Table S2 ^{31 32 33 34} Fig. 6A Figure S7 Fig. 6B Fig. 6C Fig. 6B

Genotoxic conditions leading to CENP-A delocalization and activated transcription of minor satellite repeats led to transcriptional activation of p21, indicative of a pre-senescent state (). In primary murine embryonic fibroblasts (MEF), accumulation of minor satellite transcripts also correlated with CENP-A delocalization, although with apparent different kinetics and lower magnitude that may reflect differences of cellular context and the increased sensitivity of primary cells to genotoxic stress. Indeed, a statistically significant and reproducible 2-fold increased transcription of centromeric transcripts was detectable after 4 hr of ETOP treatment together with relocation of CENP-A. These events correlated with increased expression of the senescent marker β-galactosidase (β-gal) and of cell cycle inhibitors like Cyclin-Dependent Kinase Inhibitor 2 A Ink4/p16(p16) that started being detectable after 24 hr ETOP treatment (). We thus assessed centromere integrity in a context of homogeneous populations of senescent cells, where cells are permanently arrested independently of a genotoxic stressor, while maintaining high levels of ATM signalling and stabilized p53. Cip INK4a INK4a Figure S4A Figure S8A,B ²¹

We first used MEFs as a classical model system as they rapidly lose their proliferation potential after a few passages in culture. MEFs at passage 6 (p6) stained positive for senescence-associated endogenous β-gal in more than 80% of the cells (), and exhibited higher levels of p21, p16and stabilization of p53 (). We also induced senescence in primary cells through the known Nutlin-3a-forced stabilization of p53that showed the same features as above (). Levels of centromeric transcripts in senescent MEFs-p6 or Nutlin-treated MEFs were 16 and 5-fold higher compared to the same cells at passage 2 (p2), respectively (). CENP-A was mislocalized in both cases of senescent cells (and). Compared to punctiform foci observed in control cells at p2, CENP-A formed stretched and rather fragmented signals in 40% of the MEFs-p6 cells (, middle panel). An additional 20% of the cells showed both stretched signals and partial accumulation close to the nucleolus (, lower panel). This effect was even more pronounced following Nutlin-3-forced p53 stabilization with almost 60% of the cells showing aberrant CENP-A localization (). DNA-FISH showed that centromeric architecture was drastically perturbed since, in addition to the known perturbation of pericentromeric heterochromatin foci detected by dense DAPI staining or major satellite probes, centromeric minor satellite signals were also highly disorganized and stretched in senescent cells (). Figure S8C Figure S8D ³⁵ Figure S8E Figure S9A,B Figs 7 S9C Figure S9C Figure S9C Fig. 7 Figure S9D Cip INK4a

Mouse embryonic fibroblasts (MEFs), NIH/3T3 cells and their CENP-A-SNAP derivatives were cultured and transfected as described. CENP-A–SNAP was constructed by inserting a PCR-generated murine CENP-A cDNA into pSNAPf vector (New England Biolabs) in frame with the SNAP tag. Mutation of the Ser-30 in alanine (S30A) was generated using the Q5Site-Directed Mutagenesis Kit according to the manufacturer’s recommendations (New England Biolabs). The eukaryotic expression vector containing four minor satellite repeats units under the control of the CMV promoter was generated as described previously. ^{62 5} ®

We followed previously described protocols for SNAP quench-chase-pulse or pulse-chase imaging. SNAP-tagged histones were pulse-labelled with 2 μM SNAP-cell TMR-star (New England Biolabs) for 15 min or quenched with 2 μM SNAP-cell Block (BTP, New England Biolabs) for 30 min. After quenching or pulse labelling, cells were washed twice with PBS and then incubated in complete medium for 30 min to allow excess compound to diffuse from cells. Then, cells were washed twice with PBS and incubated for 24 h in complete medium containing or not 10 μM Etoposide. Labelling of total CENP-A was then monitored by immunofluorescence as described below. ⁶³

Immunofluorescence was performed as previously described. ⁵

Cells were directly grown on Superfrost Plus microscope slides (Menzel–Glaser, Braunschweig, Germany) then fixed and permeabilized as described above. DNA-FISH and RNA-FISH were performed as previously described. ⁵

Image acquisition was performed at room temperature on a fluorescence microscope (Axioplan 2; Zeiss) with a Plan-Neofluar 100X/1.3 NA oil immersion objective (Zeiss) using a digital cooled camera (CoolSNAP fx; Photometrics) and METAMORPH 7.04 software (Roper Scientific, Trenton, NJ). Images presented correspond to one focal plane. Analysis was performed by scoring at least 150 cells in each experiment.

siRNA purchased from Sigma or Eurofins MWG Operon () were transfected into cells using Interferin (Polyplus transfection) following manufacturer’s instructions. Table S5

Total RNA from cell lines was isolated using TRI Reagent(Sigma) according to manufacturer’s instructions. Contaminant genomic DNA was eliminated with TURBO DNA-free kit (Ambion). Real-time PCR was performed using the LightCyclerDNA Master SYBR Green I mix (Roche) supplemented with 0.2 μM specific primer pairs () and analysed by the comparative CT(ΔΔCT) method using U6 RNA as an invariant RNA. Each data shown in RT-qPCR analysis is the result of at least three independent experiments performed on at least three independent RNA extractions. ® ® Table S6

Nuclear fractionation and western blot analysis was performed as before. ⁶²

ChIP was essentially performed as described. Sequences of primers are listed in. ⁶² Table S6