DNA-SCARS: distinct nuclear structures that sustain damage-induced senescence growth arrest and inflammatory cytokine secretion

Cellular senescence limits the proliferation (growth) of damaged cells that are at risk for neoplastic transformation by imposing an essentially irreversible growth arrest. Cells senesce in response to many potentially oncogenic stressors, including dysfunctional telomeres, DNA damage, chromatin alterations and strong mitogenic signals such as those delivered by some oncogenes (Ben-Porath and Weinberg, 2004; Campisi and d'Adda di Fagagna, 2007). The senescence response depends crucially on the cellular tumor antigen p53 (also known as tumor suppressor TP53) and the retinoblastoma-associated protein (pRb) tumor suppressor pathways and is now accepted as a potent cell-autonomous mechanism for suppressing the development of cancer (Braig and Schmitt, 2006; Campisi, 2005; Dimri, 2005; Prieur and Peeper, 2008). Accordingly, loss of the senescence response increases the incidence of cancer in humans and mice.

Unlike apoptotic cells, which rapidly disintegrate, senescent cells remain viable in culture for long intervals and are found with increasing frequency in aged tissues and at sites of age-related pathology, including preneoplastic lesions (Collado et al., 2005; Dimri et al., 1995; Erusalimsky and Kurz, 2005; Jeyapalan et al., 2007; Price et al., 2002). In addition, they develop a senescence-associated secretory phenotype (SASP) with potent autocrine and paracrine activities. The SASP includes numerous cytokines, growth factors and proteases, and develops several days after cells receive a senescence stimulus and cease growth (Coppe et al., 2010; Coppe et al., 2008; Rodier et al., 2009). Some SASP components reinforce the growth arrest (Acosta et al., 2008; Kuilman et al., 2008; Wajapeyee et al., 2008). Others disrupt epithelial differentiation (Parrinello et al., 2005) or promote cancer cell growth and invasion in culture and in vivo (Bavik et al., 2006; Coppe et al., 2008; Krtolica et al., 2001; Liu and Hornsby, 2007). Because senescent cells can strongly influence nearby cells, it is important to understand how the SASP develops.

Several signaling cascades are associated with the establishment and maintenance of senescence-associated phenotypes, including growth arrest and SASP (Campisi and d'Adda di Fagagna, 2007; Kuilman and Peeper, 2009). Many senescence-inducing stimuli generate a persistent DNA damage response (DDR), normally associated with DNA double-strand breaks (DSBs) (d'Adda di Fagagna, 2008). Recent findings show that DDR signaling is essential for establishing and maintaining senescent phenotypes. Thus, loss of DDR checkpoint kinases such as ATM or the serine/threonine-protein kinase CHK2, which phosphorylate and activate p53, not only prevents the p53-dependent senescence growth arrest (Bartkova et al., 2006; Beausejour et al., 2003; Di Micco et al., 2006; Gire et al., 2004; Herbig et al., 2004) but also prevents the p53-independent inflammatory cytokine secretion that comprises the SASP (Rodier et al., 2009).

DDR signaling is initiated at DSBs by sensor proteins such as the phosphoinositide 3-kinase-like kinases (PIKKs) ATM and ATR, and amplified by the MRN (MRE11–RAD50–NBS1) complex. These proteins help recruit and further activate PIKKs, and participate in DNA repair. PIKKs promote local chromatin remodeling, which spreads for megabases surrounding the DSB and facilitates repair. PIKKs also transduce the DDR signal to downstream mediators, such as CHK2 and p53, which integrate the signal with cellular physiology and coordinate DNA repair with cell cycle checkpoints (Bartek and Lukas, 2007; Berkovich et al., 2007; Rodier et al., 2007). Many of these DDR signaling and repair proteins assemble rapidly (within minutes) around DSBs and can be detected in the nuclei of fixed or living cells as focal aggregates termed DNA damage foci. Two components are typically used to detect these foci by fluorescence microscopy: the PIKK-phosphorylated form of the histone variant H2A.x (γH2AX), and the adaptor protein tumor suppressor p53-binding protein 1 (53BP1) (Celeste et al., 2003; Huyen et al., 2004; Lobrich et al., 2010; Meier et al., 2007; Rogakou et al., 1999).

When DNA lesions are repairable, DNA damage foci are transient. They typically resolve within 24 hours, during which time cells transiently arrest growth, presumably to allow time for repair. However, severe or irreparable DNA damage, such as complex breaks or uncapped telomeres, causes many cells to senesce with persistent DNA damage foci, constitutive DDR signaling and chronic p53 activation. These persistent changes precede establishment of senescence-associated phenotypes, including growth arrest (Beausejour et al., 2003; d'Adda di Fagagna et al., 2003; Herbig et al., 2004) and SASP (Coppe et al., 2008; Rodier et al., 2009).

Much is known about the hierarchy and temporal recruitment of DDR signaling and repair proteins that aggregate immediately after DNA damage and disaggregate as transient damage foci resolve (Bekker-Jensen et al., 2006; Lisby et al., 2004; Paull et al., 2000). Many proteins that associate with transient foci also occur in the persistent foci that mark senescent cells (d'Adda di Fagagna et al., 2003; Herbig et al., 2004; Kim et al., 2004; Takai et al., 2003). However, DNA damage foci per se might not be accurate senescence markers. For example, telomere dysfunction-induced foci (TIF) – γH2AX or 53BP1 foci localized to telomeres – can identify senescent cells in culture and tissues (Herbig et al., 2006; Herbig et al., 2004; Takai et al., 2003), but TIF can also occur in pre-senescent cells (Beliveau et al., 2007; Verdun et al., 2005) and the damage foci in senescent cells are not always telomere associated (Nakamura et al., 2008; Passos et al., 2010; Wang et al., 2009). Furthermore, it is not known whether the persistent foci associated with senescence contain components that are distinct from those in transient foci. In addition, while the persistent foci clearly contain active DDR components, it is not known whether these structures maintain the DDR activity that is required for the senescence-associated growth arrest or cytokine secretion.

Here, we investigate the spatiotemporal dynamics of persistent DNA damage foci and the impact these structures have on establishing and maintaining senescence-associated phenotypes. We use X-irradiation to initiate damage-induced senescence synchronously in normal human cells, follow DNA damage foci over time and identify characteristics that distinguish transient from persistent foci. Because markers of persistent foci are recruited to locally modified chromatin, we termed these persistent foci ‘DNA segments with chromatin alterations reinforcing senescence’ or DNA-SCARS. Finally, we show that DNA-SCARS are important for maintaining the senescence-associated growth arrest and secretion of IL-6, an important SASP component.

To investigate the kinetics of damage foci formation and resolution, we exposed proliferating normal human fibroblasts (strain HCA2) to 0.5 Gy ionizing radiation (IR; X-ray), which causes a transient growth arrest that reverses within 24 hours; alternatively, we used 10 Gy, which causes a (permanent) senescence arrest (Rodier et al., 2009). We monitored damage foci by 53BP1 immunofluorescence. As expected, 0.5 Gy rapidly produced many small foci, which resolved almost completely within 12 hours; only a few residual foci remained 24 hours later (Fig. 1A; supplementary material Fig. S1A). 10 Gy also rapidly produced many small foci, but a subfraction failed to resolve, became enlarged and persisted for days (Fig. 1A,E) and months (data not shown). These IR-induced persistent DNA damage foci contained many proteins that are also present in transient foci or foci that mark dysfunctional telomeres (d'Adda di Fagagna et al., 2003; Herbig et al., 2004; Lisby et al., 2004; Meier et al., 2007; Rogakou et al., 1999; Takai et al., 2003). Such proteins included the modified chromatin component γH2AX (see Fig. 1H), the repair or adaptor proteins MDC1 (see below), NBS1 and MRE11 (data not shown), the activated DDR signaling protein ATM-pSer-1981 (data not shown) and proteins containing the ATM/ATR-pSer/Thr substrate motif (supplementary material Fig. S1B). Virtually all the 10-Gy-irradiated proliferating cells developed persistent foci, demonstrating that the cell cycle phase at the time of irradiation did not influence their formation. A senescence-inducing dose of bleomycin, a radiomimetic used in cancer therapy (Regulus et al., 2007), also produced both small transient and large persistent foci (Fig. 1H, nonpermeabilized panels). Not surprising, the higher IR and bleomycin doses produced more small transient foci, and possibly some larger early foci (although difficult to distinguish from coincident foci), than lower doses, but, in all cases, the foci largely resolved after low doses, whereas large foci persisted after the higher doses.

To determine whether transient and/or persistent foci are sites of active DNA synthesis, indicative of DNA replication or repair, we exposed proliferating cells to 10 Gy IR, pulsed them with bromodeoxyuridine (BrdU) for 24 hours at varying times after IR and analyzed single cells.

In unirradiated cells, >80% showed uniform nuclear staining, consistent with passage through S phase. ~15% of unirradiated cells contained BrdU-positive foci, indicating DNA synthesis and presumably repair of spontaneous damage at these sites (Fig. 1B,C).

As expected (Rodier et al., 2009), 10 Gy IR arrested cell cycle progression within 24 hours. Thus, between 0.5 and 24 hours after IR, only ~50% of cells showed uniform BrdU labeling; this fraction declined to <30% 1.5–26 hours after IR, and <1% 24–48 hours after IR (Fig. 1B,C). DNA synthesis was also evident in foci, but only those that were pulsed early after IR. Thus, 20–25% of cells had BrdU-positive foci when pulsed 0.5–24 hours or 1.5–26 hours after IR. However, 48 hours after IR, few if any cells had BrdU-positive foci, despite numerous persistent foci (Fig. 1B,C). Even long BrdU pulses, from 3–8 days, failed to label persistent foci in irradiated senescent cells; the same was true for replicatively senescent cells, which contain persistent damage foci localized to dysfunctional telomeres (TIF) (data not shown). We conclude that persistent damage foci are not sites of replicative or repair DNA synthesis.

Because we irradiated proliferating cells, we hypothesized that cells irradiated during S phase incorporated BrdU into foci owing to homologous recombination repair (HRR), the preferred mode of DSB repair during S phase. Indeed, immunostaining revealed the presence of RPA70, a single-strand DNA (ssDNA) binding and HRR protein, in some 53BP1 foci. A fraction of RPA70 also remained nucleoplasmic. However, focal coincidence between RPA70 and 53BP1 was limited to 24 hours after IR, with a peak ~18 hours after IR (Fig. 1D,E). RPA70 and 53BP1 coincided at frequencies similar to the number of cells in S phase (compare Fig. 1C and 1E), but foci that persisted for >48 hours were devoid of RPA70. RPA70 was also absent from 53BP1 foci in replicatively senescent cells (Fig. 1E), suggesting that persistent foci, regardless of origin, lack ssDNA. We also exposed proliferating cells to 10 Gy IR, allowed recovery for 3 hours, pulsed with BrdU for 4 hours, then stained for BrdU and RAD51, another ssDNA binding or HRR protein. Cells with RAD51 foci were almost exclusively positive for BrdU (Fig. 1F,G). Thus, in contrast to early transient foci, the persistent foci that remain 24–48 hours after senescence-inducing damage lack evidence of DNA synthesis, ssDNA and HRR.

All proliferating cells exposed to 10 Gy developed persistent foci, suggesting that S phase (or repair associated with S phase) is not required for their formation or maintenance. To test this idea, we arrested cell proliferation using a lentivirus to overexpress the cyclin-dependent kinase inhibitor 2A p16^INK4A. As expected, cells arrested with a G1 DNA content ~24 hours after infection. Forty eight hours after infection, we irradiated the cells with 10 Gy. Dual immunostaining showed that the p16^INK4A-arrested cells resolved early transient foci in a manner similar to that of pre-senescent cells. Thus, at least for cells induced to senesce by p16^INK4A, early repair is not strongly affected, although other senescent cells might repair DSBs less efficiently (Gorbunova et al., 2007). Importantly, 5 days later, most irradiated p16^INK4A-positive cells harbored persistent 53BP1 foci, whereas few unirradiated p16^INK4A-positive cells had foci (supplementary material Fig. S1C,D). Because persistent foci do not incorporate BrdU or harbor certain repair proteins (RPA70, RAD51) found in early foci, we suggest that they might not be sites of damaged DNA per se but instead are stable chromatin alterations resulting from damage. Indeed, live-cell imaging of damage foci labeled with MDC1–eGFP showed that a significant fraction of the persistent foci were stable during at least 6 hours (supplementary material Fig. S1E), a period that was largely sufficient to resolve most of the transient foci induced by 0.5 Gy IR (Fig. 1A; supplementary material Fig. S1A). Similar observations were recently made by others using a 53BP1–eGFP fusion protein (Passos et al., 2010). For this and the reasons explained below, we refer to persistent foci as DNA-SCARS.

To evaluate how proteins associate with transient foci and DNA-SCARS, we treated cells with bleomycin and either fixed before permeabilization to detect both soluble and non-soluble proteins or permeabilized before fixation to extract soluble proteins and detect only non-soluble proteins. Pre-permeabilization completely removed RPA70 from the nucleus (Fig. 1H, upper panels), indicating that RPA70 is readily soluble, regardless of whether it is nucleoplasmic or in early foci. By contrast, pre-permeabilization failed to remove γH2AX (lower panels), regardless of whether it was in early foci or DNA-SCARS (Fig. 1H), as expected for a core histone tightly integrated into chromatin. 53BP1, however, extracted differentially, depending on the type of foci. Pre-permeabilization removed 53BP1 from foci that formed early after damage but failed to remove 53BP1 from DNA-SCARS. This difference in extractability could be due to the greater amount of 53BP1 in DNA-SCARS (Fig. 1H) but nonetheless identifies a notable difference between early transient foci and DNA-SCARS. This finding supports the idea that early foci differ from DNA-SCARS, one characteristic being decreased 53BP1 solubility.

In many human cells, active p53 is required for both establishment and maintenance of the senescence growth arrest (Beausejour et al., 2003; Gire and Wynford-Thomas, 1998; Rodier et al., 2009), suggesting that there is a reservoir of active p53 in senescent cells. p53 activation occurs through stabilization and posttranslational modifications, particularly Ser15 phosphorylation (p53-pS15) by DDR kinases such as ATM (Appella and Anderson, 2001). After 10 Gy IR, HCA2 cells showed the expected rapid rise in p53 and p53-pS15 levels, followed by a decline, as detected by western analysis (Fig. 2A). However, p53 and p53-pS15 typically declined to levels slightly above those of pre-senescent controls (Fig. 2A). This finding raised the possibility that DNA-SCARS, which contain activated ATM (d'Adda di Fagagna et al., 2003; Herbig et al., 2004; Rodier et al., 2009), might also harbor activated p53 and act as reservoirs to maintain its activity.

Depending on the type of cell and damage, p53 has been reported present in (Al Rashid et al., 2005) or absent (Bakkenist and Kastan, 2003; Bekker-Jensen et al., 2006) from early DNA damage foci. We therefore asked whether p53 was present in early foci or DNA-SCARS in normal human fibroblasts. Both p53 and p53-pS15 appeared in situ at 53BP1 foci within 5 hours of 10 Gy IR; moreover, both remained associated with DNA-SCARS for at least 10 days after IR and in replicatively senescent (SnR) cells (Fig. 2B,C). The specificities of the p53 antibodies were confirmed by immunostaining bleomycin-damaged human cancer cells that were either wild-type or null for p53 (supplementary material Fig. S1F).

These results demonstrate that at least a fraction of the low level of p53-pS15 detected in senescent cells by western analysis (Fig. 2A) is contained in DNA-SCARS, suggesting that they serve as reservoirs for the p53 activity that maintains the growth arrest.

CHK2 is a downstream DDR effector that phosphorylates p53 and is required to maintain a senescence growth arrest (Gire et al., 2004). Activated CHK2 (phosphorylated on threonine 68, pT68) was reported absent from early damage foci in human cancer cells (Lukas et al., 2003) but present at TIF in normal replicatively senescent cells (Herbig et al., 2004). We immunostained early damage foci and DNA-SCARS in normal human fibroblasts for CHK2-pT68. Minutes after 10 Gy IR, CHK2-pT68 staining intensity increased, but it did so throughout the nucleus and without discernible enrichment at foci (Fig. 2D). This nucleoplasmic staining generally declined within 24 hours, after which focal staining began to appear. Many hours later, CHK2-pT68 clearly localized to DNA-SCARS (Fig. 2D). The specificity of the phospho-CHK2 antibody was confirmed by immunostaining irradiated wild-type and CHK2-null cancer cells (supplementary material Fig. S1G) and using a different antibody against phospho-CHK2 (supplementary material Fig. S1H). We also detected CHK2-pT68 in the DNA-SCARS (TIF) of replicatively senescent cells (data not shown). These results reconcile earlier results (Herbig et al., 2004; Lukas et al., 2003) and suggest that, although activated CHK2 diffuses through the nucleus immediately after damage, a fraction accumulates at DNA-SCARS in senescent cells. They also suggest that TIF are selectively localized DNA-SCARS.

PML (promyelocytic leukemia protein) defines a dynamic, heterogeneous subnuclear domain (nuclear body, NB) that participates in cellular responses to stress, including genotoxic stress (Bernardi and Pandolfi, 2007). PML NBs can be found at DNA damage foci (Carbone et al., 2002; Xu et al., 2003) and act as general sensors of genomic damage (Varadaraj et al., 2007). Because PML also facilitates p53 activation and the senescence growth arrest (Ferbeyre et al., 2000; Pearson et al., 2000), we evaluated its relationship to damage foci during the transition to senescence after IR.

Senescence-inducing IR (10 Gy) generated 53BP1 foci that initially showed no preferential localization to PML NBs (Fig. 3A). However, as DNA-SCARS formed, most of them associated with a PML NB. This association began ~24 hours after IR, was complete 48 hours later (Fig. 3A) and persisted for many days (Fig. 3B; note, ~10% of control cells had damage foci, most of which were PML associated (70%), suggesting that they were DNA-SCARS or TIF in the senescent cells that are present in most normal human cell populations). Unlike 53BP1 and γH2AX, which completely colocalize, PML NBs were often at the periphery of DNA-SCARS. This was true for IR-induced (Fig. 3A,B), bleomycin-induced and replicative senescence (supplementary material Fig. S2A,B).

The association between DNA-SCARS and PML NBs was specific and did not occur with other nuclear structures such as centromeres or telomeres. Twenty four hours after 10 Gy IR, when many 53BP1 foci (DNA-SCARS) associated with PML NBs, they were largely excluded from centromeres, as determined by staining for CREST (Fig. 3A), and telomeres, as determined by staining for TRF2 (Fig. 3A). Because PML NB association began ~24 hours after IR, we asked whether RPA70, which showed a peak of association with damage foci ~18 hours after IR, was excluded from PML-associated foci. Triple immunofluorescence showed that, 24 hours after IR, before RPA70 foci completely declined, it was possible to detect foci containing RPA70, 53BP1 and PML (Fig. 3C). Thus, ssDNA-binding proteins do not preclude association of DNA-SCARS with PML NBs.

The formation and resolution of damage foci are linked to damage signaling and repair pathways. We therefore asked whether important components of these pathways altered DNA-SCAR formation, as determined by association with PML NBs.

To test the requirement for p53, we infected proliferating HCA2 cells with a retrovirus encoding GSE22, an often-used peptide that stabilizes p53 but prevents formation of active tetramers (Ossovskaya et al., 1996). These cells develop spontaneous damage foci after 5–10 doublings (Rodier et al., 2009). Despite loss of p53 function, 53BP1 foci associated with PML NBs, suggesting that they are DNA-SCARS and do not require p53 for formation (Fig. 4A). Similar results were obtained with cells infected with a retrovirus encoding the viral oncogene E7, which inactivates pRb, or a lentivirus encoding SV40LT, which inactivates p53 and pRb (Fig. 4A). Thus, formation of DNA-SCARS did not depend on functional p53 or pRb pathways. Importantly, cells deficient for either p53 or pRB harbored senescence levels of DNA-SCARS but proliferated, indicating that DNA-SCARS can be uncoupled from the senescence growth arrest when cell cycle checkpoints are defective.

We also examined human fibroblasts deficient for the ATM or ATR signaling kinases, the Artemis repair-associated nuclease, the NBS1 DNA damage signaling transducer or the BLM HRR helicase. All these cells developed DNA-SCARS (53BP1 foci associated with PML NBs) ~24 hours after 8 Gy IR. Thus, DNA-SCARS did not require ATM, ATR, Artemis, NBS1 or BLM for formation. Furthermore, cells deficient in NBS1, BLM or Artemis formed DNA-SCARS faster than wild-type cells (Fig. 4B), suggesting that DNA-SCARS might result from faulty DNA repair.

We also asked whether PML NBs were essential for DNA-SCARS formation or maintenance. We used lentiviruses to express eGFP (control), the PML-RAR oncogene, or the human cytomegalovirus protein iE1 (also known as ICP0) in human fibroblasts. PML-RAR disperses PML into microspeckles (Fig. 4C) and is a dominant inhibitor of wild-type PML function (Dyck et al., 1994). iE1 prevents PML sumoylation, causing diffuse nucleoplasmic organization (Fig. 4C), an absence of PML NBs and loss of PML function (Ahn and Hayward, 1997). We irradiated the infected cells and followed the appearance, resolution and retention of 53BP1 foci over time. The rapid appearance and resolution of early 53BP1 foci after 0.5 or 10 Gy IR were unaffected by PML-RAR or iE1 (Fig. 4D). Likewise, the appearance of DNA-SCARS was unaffected by PML-RAR or iE1 (Fig. 4D). Thus, although DNA-SCARS associate with PML-NBs, PML function is not required for formation of DNA-SCARS.

We previously showed that persistent damage foci correlate with the DDR signaling that is essential for maintaining the senescence-associated growth arrest and inflammatory cytokine secretion (Rodier et al., 2009), consistent with our finding here that DNA-SCARS contain active CHK2 and p53 (Fig. 2). Histone H2AX is dispensable for the initial formation of DNA damage foci but is required for their stabilization (Celeste et al., 2003). We therefore tested the idea that depletion of H2AX might disrupt DNA-SCARS and hence senescence-associated phenotypes.

We depleted H2AX by RNA interference (RNAi) using lentiviruses encoding short-hairpin (sh) RNAs against eGFP (control) or H2AX. Western blotting confirmed the decrease in levels of H2AX (Fig. 5A). We then followed the formation of 53BP1 foci after wild-type and H2AX-deficient cells were irradiated with 10 Gy. Forty eight hours after IR, 53BP1 efficiently formed persistent foci in both cell types. However, the DDR adaptor protein MDC1 and CHK2-pT68 were either absent or present at much reduced levels in H2AX-depleted 53BP1 foci (Fig. 5B). Thus, H2AX deficiency interfered with the efficient assembly of some DDR proteins into DNA-SCARS.

To determine the functional consequences of H2AX deficiency, we irradiated (10 Gy) control and H2AX-deficient cells. After 7 days, we pulsed the cells for 24 hours with BrdU to assess DNA synthesis. H2AX deficiency increased BrdU incorporation by cells that would normally remain growth arrested (Fig. 5C,D). All four H2AX shRNAs tested gave similar results. We also asked whether H2AX-deficient cells secreted a SASP cytokine in response to senescence-inducing DNA damage (Rodier et al., 2009). H2AX deficiency suppressed IL-6 secretion 2–3 days after IR, when cells just begin secreting, and 9–10 days later, when secretion is more robust (Fig. 5E). Thus, H2AX depletion reduced certain DDR signaling proteins in DNA-SCARS and suppressed the DDR-dependent senescence-associated growth arrest and IL-6 secretion.

We tested the occurrence of DNA-SCARS, determined by association of 53BP1 foci with PML NBs, in HCA2 cells induced to senesce by means other than IR: expression of an activated oncogene (RAS) (Serrano et al., 1997), expression of a dominant-negative telomeric protein (DN-TIN2) (Kim et al., 2004) or treatment with hydrogen peroxide (H₂O₂). Each manipulation induced DNA-SCARS within 48 hours (Fig. 6A). Furthermore, we observed DNA-SCARS in four different human fibroblast strains (WI-38, IMR-90, BJ, HCA2) induced to senesce by replicative exhaustion or IR (data not shown) and in IR-induced senescent human mammary epithelial cells (HMECs), human umbilical vein endothelial cells (HUVECs) and human aortic endothelial cells (HAECs) (Fig. 6B). These findings suggest that DNA-SCARS are a general feature of damaged senescent cells.

To determine whether DNA-SCARS form in vivo, we examined mouse cells and tissues. First, we exposed primary mouse embryonic fibroblasts (MEFs) to 10–100 Gy IR and, 8 days later, immunostained for 53BP1 and PML. Coincident staining was readily apparent, although higher IR doses were needed relative to those in human cells (Fig. 7A,B). Thus, DNA-SCARS form in both human and mouse cells. We then subjected mice to whole-body non-lethal IR (8 Gy) and prepared frozen tissue sections for immunostaining at varying times thereafter. 53BP1 foci were readily detected in both the alveoli and bronchioles of lungs 6 hours, 48 hours and 7 days after IR (Fig. 7C,D; supplementary material Figs S3, S4). Nearly 100% of cells harbored foci 6 hours after IR. Many foci resolved after 48 hours, especially in the bronchioles, but a fraction of cells (45–70%) maintained at least one 53BP1 focus 1 week after IR (Fig. 7D; see supplementary material Fig. S5 for similar results in liver). After 14 weeks, persistent 53BP1 foci diminished further, but remained two- to four-fold higher than the low level in matched unirradiated controls (Fig. 7D). This delayed decline might reflect damage repair, clearance of damaged cells by the immune system (Ventura et al., 2007; Xue et al., 2007) and/or cell turnover (Le et al., 2010).

To determine whether the 53BP1 foci that persisted 1 week after IR were DNA-SCARS, we asked whether they associated with PML NBs. Fig. 7E illustrates how bronchiolar layers were selected and single cells analyzed for coincidental 53BP1 and PML staining. As shown in the Fig. 7F insets, most bronchiolar epithelial cell nuclei showed an association between 53BP1 and PML NBs (cells 1–4). Additionally, non-epithelial cells (underlying the airway epithelial border) also showed associated 53BP1 foci and PML NBs (cells 5–6). Similar results were obtained in liver and lung alveolar cells, although only 30–50% of 53BP1 foci associated with PML NBs in these compartments (supplementary material Figs S4, S5). These results suggest that association between PML NBs and 53BP1 can be used to distinguish early transient DNA damage foci from DNA-SCARS in vivo.

Constitutive DDR signaling is an important component of the integrated tumor suppressor network that establishes and maintains the senescence phenotypes of growth arrest and the SASP (d'Adda di Fagagna et al., 2003; Herbig et al., 2004; Rodier et al., 2009). Here, we show that senescence-associated DNA damage foci, or DNA-SCARS, are relatively stable structures that are distinct from transient damage foci and functionally important for both the growth arrest and IL-6 secretion, an important SASP feature.

DNA damage foci can be detected by immunofluorescence because chromatin modifications generated by the DDR at DSBs spread over megabases (Meier et al., 2007; Rogakou et al., 1999). Whether these foci form early and transiently after DNA damage or are the persistent foci associated with senescence, they reflect elevated concentrations of modified histones (e.g. γH2AX) and other proteins that are recruited to remodeled chromatin. We identified several molecular events that differentiate transient from persistent foci, which we followed after X-irradiation to generate DSBs (among other lesions) synchronously.

In transient foci, lesions are likely to be repaired by both non-homologous end joining (NHEJ) and HRR, depending on the cell cycle phase in which the damage was inflicted. For cells damaged in S phase, we detected the ssDNA-binding proteins RAD51 and RPA70, which bind the resected DNA that is generated during HRR; these cells also incorporated BrdU, indicative of repair. γH2AX can be found at ssDNA, in addition to its predominant association with DNA DSBs (Lobrich et al., 2010). Nonetheless, both ssDNA-binding proteins and DNA synthesis were absent from persistent foci. Furthermore, individual persistent foci were stable for several hours in living cells, and, in fixed senescent cells, were detected for days and weeks after their formation. Thus, HRR and NHEJ are likely either inactive or incapable of resolving these foci after ~48 hours. Formation of DNA-SCARS was accelerated in cells that are deficient in certain DNA repair proteins (Fig. 4B), supporting the idea that ineffective or defective repair initiates the formation of these structures. We propose that, by ~48 hours after senescence-inducing DNA damage, a general repair phase ends and the remaining foci contain either unresolved lesions or stable chromatin modifications resulting from the damage. Alternatively, repair mechanisms that do not require DNA synthesis, HRR or NHEJ could be actively attempting to repair lesions in DNA-SCARS, but these (presumably abortive) repair processes were not detected in our study. Because these foci persist, and their destabilization relaxed the senescence growth arrest and IL-6 secretion, we term them ‘DNA segments with chromatin alterations reinforcing senescence’ or DNA-SCARS.

DNA-SCARS include TIF, which occur at dysfunctional telomeres (Herbig et al., 2004; Takai et al., 2003). They also include foci generated by senescence-inducing DNA damage inflicted by H₂O₂ or oncogenic RAS. However, as with other senescence markers, DNA-SCARS are associated with, but not exclusive to, senescent cells. In particular, p53-deficient cells spontaneously develop DNA-SCARS, yet they proliferate. Such cells resemble those in pre-cancerous lesions in which DDR signaling is detected before full transformation (Gorgoulis et al., 2005). Furthermore, PML NBs can harbor damaged telomeres in telomerase-deficient cancer cells (Nabetani et al., 2004). In these ALT (‘alternative lengthening of telomeres’) cells, markers of telomeres, DSBs and PML colocalize in a structure reminiscent of DNA-SCARS. These results suggest that, although DNA-SCARS are novel markers of senescence in normal cells, the association can be uncoupled in preneoplastic or cancer cells. Thus, DNA-SCARS or TIF can be used as senescence markers, but only when combined with other markers, such as senescence-associated beta-galactosidase (SA-Bgal) (Dimri et al., 1995), p16^INK4A (Krishnamurthy et al., 2004) or senescence-associated heterochromatic foci (Narita et al., 2003). Notably, only the growth arrest, not inflammatory cytokine secretion, was uncoupled from DNA-SCARS in cells expressing viral oncogenes that inactivate the p53 and/or pRB tumor-suppressor pathways. Therefore, in the context of cancer, DNA-SCARS could mark locally increased inflammatory cytokine secretion, which is associated with persistent DDR signaling (Rodier et al., 2009).

The association between DNA-SCARS and PML NBs, where many repair proteins localize, might occur to allow further processing of lesions. In addition, persistent foci might associate with PML NBs, where many chromatin-modifying proteins also localize, to promote senescence-associated gene expression changes such as the SASP. Alternatively, the presence of repair proteins in PML NBs could simply reflect their normal localization when the immediate repair phase ends. In support of this idea, disruption of PML NB structures did not alter the kinetics of foci formation and resolution. Furthermore, loss of several DDR repair and checkpoint proteins did not prevent, and in some cases accelerated, DNA-SCAR formation. Thus, damage foci might quickly progress to DNA-SCARS in the absence of optimal repair.

Senescent cells acquire a complex phenotype, including arrested growth and a multi-factor SASP (Coppe et al., 2008). We propose that DNA-SCARS provide a reservoir for active DDR signaling, which is essential to maintain both the p53-dependent growth arrest and inflammatory cytokine secretion (Beausejour et al., 2003; d'Adda di Fagagna et al., 2003; Gire et al., 2004; Gire and Wynford-Thomas, 1998; Herbig et al., 2004; Rodier et al., 2009). In support of this idea, depletion of the damage-foci-stabilizing core histone H2AX disrupted the DNA-SCARS structure, reducing the levels of SCARS-associated MDC1 and activated CHK2 and relaxed both the senescence growth arrest and IL-6 secretion. Although ATM and CHK2 are essential for the senescence-associated secretion of inflammatory cytokines (Rodier et al., 2009), p53 is not, suggesting that DNA-SCARS could be the reservoir for active p53, CHK2 and other activated DDR proteins. Thus, DNA-SCARS influence two important features of the senescent phenotype – growth arrest and inflammatory cytokine secretion.

DNA-SCARS appeared after a senescence-inducing dose of DNA damage in all primary human and mouse fibroblasts studied, in primary human epithelial and endothelial cells and in mouse tissues. This finding indicates that formation of DNA-SCARS is a robust phenomenon that is conserved between mouse and human, as are many other features of cellular senescence (Coppe et al., 2010), and that it occurs in culture and in vivo. A deeper understanding of how these structures assemble and function will probably enrich our insights into the mechanisms that link DNA damage, inflammation, cancer and aging. Furthermore, it might suggest novel strategies for ameliorating the effects of DNA damage, whether encountered during normal aging, environmental exposure or certain therapies, such as those used to treat many cancers.

In general, cells were cultured at 37°C in 10% CO₂, DMEM, 10% fetal bovine serum and 100 U/ml streptomycin–penicillin. Early-passage cells had 24 hours BrdU labeling indices of >75%, and <10% expressed SA-Bgal (Dimri et al., 1995). Cell populations were considered senescent when they had labeling indices of <5% and >75% expressed SA-Bgal. HCA2 cells were from J. Smith (University of Texas, San Antonio, TX). Artemis-deficient cells were from S. Yannone, ATM-deficient cells from E. Blakeley and human mammary epithelial cells from M. Stampfer, all from Lawrence Berkeley National Laboratory. Fibroblasts deficient in ATR (GM09812), NBS1 (GM07166A) or BLM (GM02085) were from the Coriell Institute. Human umbilical vein endothelial cells (HUVECs) and aortic endothelial cells (HAECs) were from CloneTics.

Viruses were produced as described previously (Beausejour et al., 2003) and titers adjusted when necessary to achieve ~90% infectivity. Retroviruses encoding eGFP (Rodier et al., 2009), GSE22 (Itahana et al., 2002) or E7 (Itahana et al., 2003) have been described previously. iE1 and PML-RAR were cloned into the pBABE-puro retroviral backbone. Lentiviruses encoding dominant-negative TIN2 (TIN2-15C) (Kim et al., 2004), eGFP (Rodier et al., 2009), SV40LT (Beausejour et al., 2003) and oncogenic RAS^V12 (Rodier et al., 2009) have been described previously. The full-length coding sequence for human MDC1 was fused with eGFP (entry vector 770-7) and inserted in the lentiviral backbone 775-1 to generate the pLenti CMV/TO GFP–MDC1 expression vector (779-2) (Campeau et al., 2009). Lentiviruses encoding shRNAs against eGFP and H2AX were purchased from Open Biosystems (shGFP3: #RHS4459; shH2AX-A: #TRCN0000073278; shH2AX-B: #TRCN0000073280; shH2AX-C: #TRCN0000073279; and shH2AX-D: #TRCN0000073281).

Cells were X-irradiated at rates equal to or above 0.75 Gy/minute using a Pantak X-ray generator (320 kV/10 mA with 0.5 mm copper filtration). Cells were treated with bleomycin, either 10 μg/ml for 30 minutes with a recovery period of 30 minutes or 20 μg/ml for 2 hours with a recovery period of 10–12 days.

C57BL6 mice were sacrificed and excised organs were dipped in OCT (Tissue-Tek #4583), flash-frozen in liquid nitrogen and preserved at −80°C. Organs were sectioned (6–10 μm thickness) at −20°C and slices were placed on superfrost plus microslides (VWR #48311-703). Slices were air-dried and stored at −80°C.

Soluble proteins were extracted as described previously (He et al., 1990). Briefly, cells in chamber-slides were washed on ice with phosphate-buffered saline (PBS) and soluble proteins extracted by an incubation for 10 minutes with ice-cold permeabilization buffer containing inhibitors of RNAses (New England Biolabs), phosphatases (Sigma) and proteases (Complete-Mini Roche). Cells were washed in cold permeabilization buffer and PBS before being fixed in formalin (room temperature) and processed for immunofluorescence (below).

Frozen tissue-section slides were brought to room temperature for 10 minutes, or cells in four-well chamber-slides were washed in PBS. Frozen sections or cells were fixed in formalin (10%, neutral buffered) for 10 minutes at room temperature, washed in PBS, permeabilized in PBS plus 0.5% Triton X-100 for 10 minutes and washed again. Slides were blocked for 1 hour in PBS plus 1% BSA and 4% serum corresponding to the secondary antibody (goat or donkey) and incubated with primary antibodies diluted in blocking buffer overnight at 4°C. The slides were washed with PBS and incubated with secondary antibodies for 1 hour at room temperature, washed and mounted with slow-fade gold (Molecular Probes). When noted, DAPI was added to secondary antibodies.

Cells in chamber slides were pulsed with BrdU before fixation and permeabilization, as described above. After two PBS washes and a wash in enzyme reaction buffer [0.75X EXO III reaction buffer (Promega) plus 10 mM MgCl₂], DNA was denatured by incubation for 30 minutes at 37°C in enzyme reaction buffer plus 5 U/ml DNASE I (Roche) and 200 U/ml EXOIII (Promega). Cells were washed with PBS and processed for immunofluorescence.

Images were acquired on an Olympus BX60 fluorescence microscope with the spotfire 3.2.4 software (Diagnostics Instruments) or a Nikon PCM-2000 laser confocal scanning microscope and prepared for figure creation using Photoshop CS2 (Adobe).

Cells were lysed in 50 mM Tris–HCl pH 6.8, 100 mM DTT, 2% SDS and 10% glycerol. The mixture was sonicated (30 seconds), boiled (10 minutes) and spun. Samples were separated by electrophoresis on 5% or 4–15% gradient Tris–glycine SDS-polyacrylamide gels. Proteins were electroblotted overnight onto PVDF membranes and probed overnight at 4°C in PBS with 5% milk. Secondary antibodies conjugated to horseradish peroxidase were used at 1:15,000 (Jackson ImmunoResearch). Blots were developed using the supersignal pico reagent (PIERCE).

The antibodies used were: 53BP1, Bethyl, BL182 or Upstate, BP13 (against human) or Novus, 100-305 (against mouse); γH2AX, Upstate, JBW301; PML, Santa Cruz Biotechnology, N-19 (against human) or Upstate, 36.1-104 (against mouse); p53, Oncogene Research Products, DO-1; phospho-p53 serine 15, Cell Signaling, #9286; p21, BD Biosciences, 556430; p16^INK4A, Neomarkers, JC8; actin, Chemicon, MAB3128; tubulin, Sigma, T5168; phospho-ATM/ATR STK substrates, Cell Signaling #2851; TRF2, Imgenex, IMG-124; CREST, Antibodies Incorporated, 15-235-F; phospho-CHK2 threonine 68, Cell Signaling #2661 (lots #1 and #7) or the rabbit monoclonal #2197; BrdU; Abcam, ab1893 or Roche, BMC 9318; Rad51, Santa Cruz Biotechnology, H-92; RPA70, Calbiochem, NA13; SV40LT, Santa Cruz Biotechnology, Pab 101; E7, Neomarkers, TVG701Y; HSP60, Santa Cruz Biotechnology, N-20; MDC1, rabbit polyclonal (Campeau et al., 2009); donkey secondary antibodies conjugated to Alexa Fluor dyes were from Molecular Probes (Alexa 350, 488 and 594).

Conditioned media (CM) were prepared by washing cells three times in PBS and incubating in serum-free DMEM plus antibiotics for 24 hours. CM were filtered and stored at −80°C. Cell number was determined for every sample. ELISAs were performed using kits and procedures from R&D (IL-6 #D06050). The data were normalized to cell number and reported as 10⁻⁶ pg secreted protein per cell per day.

We thank Patrick Kaminker for valuable assistance with the confocal microscope, Dolf Beems for comments on mouse histology and Pierre Desprez for critically reading the manuscript. This work was supported by grants from the Canadian Institutes of Health Research MPO79317 to C.B., US NIH AG09909 and AG17242 to J.C. and AG025708 to the Buck Institute, and DOE contract AC03-76SF00098. Deposited in PMC for release after 12 months.