The Tumor Suppressor PML Specifically Accumulates at RPA/Rad51-Containing DNA Damage Repair Foci but Is Nonessential for DNA Damage-Induced Fibroblast Senescence

Abstract

The PML tumor suppressor has been functionally implicated in DNA damage response and cellular senescence. Direct evidence for such a role based on PML knockdown or knockout approaches is still lacking. We have therefore analyzed the irradiation-induced DNA damage response and cellular senescence in human and mouse fibroblasts lacking PML. Our data show that PML nuclear bodies (NBs) nonrandomly associate with persistent DNA damage foci in unperturbed human skin and in high-dose-irradiated cell culture systems. PML bodies do not associate with transient γH2AX foci after low-dose gamma irradiation. Superresolution microscopy reveals that all PML bodies within a nucleus are engaged at Rad51- and RPA-containing repair foci during ongoing DNA repair. The lack of PML (i) does not majorly affect the DNA damage response, (ii) does not alter the efficiency of senescence induction after DNA damage, and (iii) does not affect the proliferative potential of primary mouse embryonic fibroblasts during serial passaging. Thus, while PML NBs specifically accumulate at Rad51/RPA-containing lesions and senescence-derived persistent DNA damage foci, they are not essential for DNA damage-induced and replicative senescence of human and murine fibroblasts.

INTRODUCTION

Cellular senescence was first observed by Hayflick in primary human cell culture systems in vitro. The major hallmark of senescent cells is a stable cell cycle arrest after a finite number of in vitro duplications (1). Senescence can be triggered by telomere shortening and nontelomeric pathways, including oncogene activation and persistent DNA damage. The pathways involving p53 and p21, as well as pRB and p16, are essential for a functional telomeric and nontelomeric DNA damage response (DDR) (2). It is now firmly established that cellular senescence can act in vivo as an important barrier against cancer progression but also contributes to aging-related tissue pathologies (3).

The finding that senescence-associated DNA damage foci (SDF) of telomeric and nontelomeric origin accumulate in senescing cells indicated DNA double-strand breaks (DSBs) as a critical factor in the senescence and aging process (4–7). Stress-induced premature senescence (SIPS) is considered to be elicited by widespread nontelomeric DNA damage in cells exposed to genotoxic stress (8). Regardless of the origin, SDF display as persistent DNA damage foci at which many known DDR factors, such as γH2AX, ATM, ATR, 53BP1, and the MRN complex, are accumulated (9). The accumulation of persistent DNA damage foci is a common process in mammalian aging in vivo and in cell culture systems (7, 10–14).

Transient foci represent sites of successful DSB rejoining, whereas persistent (late) foci contain unrepairable DSBs (7, 15, 16). The two types of foci can also be distinguished by their DNA repair protein contents (7) and spatial association with PML nuclear bodies (15, 17–19). More recently, it was shown that persistent foci lack evidence of DNA synthesis, single-stranded DNA (ssDNA), and homologous recombination repair (19).

PML nuclear bodies (NBs) are spherical protein accumulations present in most mammalian cell nuclei (20). Their major structural component is PML. Some factors interacting with PML are linked to the DDR, and therefore, PML bodies are proposed to be involved in DNA repair, apoptosis, cellular senescence, and tumor suppression (21–25). PML NBs were also found in spatial proximity to DNA single-strand breaks (SSBs) and DSBs (17, 26, 27). This suggests that PML NBs could serve as DNA damage sensors, DNA repair compartments, and physical sites where DNA repair activities and/or cell cycle checkpoint pathways are coordinated and monitored (17, 28). PML protein levels and the number of NBs are elevated when cells encounter stress, e.g., after DNA damage (28–30) and during senescence induction (31, 32).

Overexpression of PML protein isoform IV induces senescence in primary human and murine fibroblasts, and this process is dependent on p53 and pRb (32–34). The underlying mechanism involves a PML VI-mediated inhibition of E2F target gene expression, followed by a proliferation block, DNA damage induction, and senescence (35). PML-depleted cells show alterations in their responses to DNA damage and senescence induction. Certain cell types from PML knockout (KO) mice showed a decreased apoptosis rate in response to multiple stimuli, including gamma irradiation (γ-IR), UV, and DNA-damaging agents (36–41). PML knockout and knockdown murine embryonic fibroblasts (MEFs) are resistant to Ras-induced senescence (40, 42). Also, the activation of p53 is reduced in PML-depleted mouse and human cells (37, 39, 40, 42, 43).

Despite all these data, the precise function of PML in the DNA damage response is not fully understood. We therefore analyzed the DNA damage response and cellular senescence in the presence or absence of endogenous PML. Surprisingly, ablation of PML did not alter the cellular DNA damage response or senescence induction in primary human or mouse embryonic fibroblasts. These observations reveal a nonessential role for PML in fibroblast senescence.

MATERIALS AND METHODS

Cell culture.

WI-38 cells were obtained from the American Tissue Culture Collection (ATCC) and were cultured in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal calf serum (FCS) in a 10% CO₂ atmosphere at 37°C. Primary human foreskin fibroblasts (HFFs) with stable, small interfering RNA (siRNA)-mediated knockdown of PML, as well as the two control cell lines, were a kind gift of Thomas Stamminger and Nina Tavalai, Erlangen, Germany. The cells were generated as described previously (44). Primary MEFs were isolated at embryonic day 13.5 (E13.5) and cultured in DMEM supplemented with 10% FCS and 200 mM l-glutamine in a 5% CO₂ atmosphere at 37°C.

Gamma irradiation and drug treatment.

Cells were gamma irradiated (¹³⁷Cs; 1 Gy/min; Gammacell GC40; Nordion, Ottawa, Canada) or UVA irradiated (data not shown). For drug-induced senescence, cells were treated constantly for 6 days with 50 μM 5-bromodeoxyuridine (BrdU) and10 μM distamycin A (DMA) as described previously (45).

UVA microbeam-induced DNA damage.

For laser damage induction, the pulsed UVA laser was coupled into a confocal laser scanning microscope (LSM 510) via the epifluorescence illumination path. A laser microbeam was focused into the middle of the field of view by a 100×, 1.3-numerical-aperture (NA) Plan Neofluar oil immersion objective (Zeiss). The UVA laser was a frequency-tripled Nd:YLF laser (Spectra Physics) delivering 20-ns-duration pulses at 350 nm with user-defined energies from 1 μJ to 200 μJ at user-defined repetition rates of 1 Hz to 1,000 Hz. Before entering the microscope, laser pulse energy was reduced by 80% with the gradient position-dependent attenuator (Laseroptik). The cells were irradiated in a quadrangular 500-μm by 535-μm area by moving the motorized x,y table, which was driven by an MCU 26 controller (Zeiss) at a speed of 1,470 μm/s. In this case, we used a 350-Hz repetition rate. By selected irradiation options, approximately 100 cells were irradiated in 25 s, with single pulses hitting the cell nucleus every 4 μm. No sensitization was used.

Indirect immunofluorescence and Western blot analysis.

Cells grown on glass coverslips were fixed in 4% formaldehyde for 10 min and permeabilized with 0.25% Triton X-100 for 3 min. Human skin samples were a kind gift of Johannes Norgauer, Jena, Germany, and have been described previously (45). Twelve-micrometer sections were fixed in 4% formaldehyde for 10 min and permeabilized with 0.25% Triton X-100 for 10 min. After the staining, coverslips were mounted in Prolong Gold antifade mounting medium with DAPI (4′,6-diamidino-2-phenylindole) (Invitrogen, Karlsruhe, Germany). For Western blotting, whole-cell extracts were prepared by incubation at 100°C for 10 min, separated by SDS-PAGE, and transferred to a nitrocellulose membrane. Quantification of band intensities was carried out with Metamorph software (MDS Analytical Technologies, Sunnyvale, CA).

Antibodies.

The primary antibodies used in murine and human cells were as follows: anti-phospho-H2AX (Ser139) (γH2AX) clone JBW301 (Millipore, Billerica, MA) and anti-β-actin A5441 and anti-α-tubulin T9026 (both from Sigma, Taufkirchen, Germany). The primary antibodies used in human cells were as follows: anti-PML and anti-SP100 (both from Peptide Specialty Laboratories, Heidelberg, Germany); anti-Mdc1 NB100-395, anti-53BP1 NB100-305, and anti-phospho-Nbs1 (Ser343) NB100-284 (all from Novus Biologicals, Littleton, CO); anti-TRF2 (IMG-124; Imgenex, San Diego, CA); anti-phospho-ATM (Ser1981) ab81292 and anti-Ki-67 ab15580 (both from Abcam, Cambridge, United Kingdom); anti-p16 BD554079 and anti-pRb BD554136 (both from BD Pharmingen, Heidelberg, Germany); anti-p21 H-164 (Santa Cruz Biotechnology, Santa Cruz, CA); anti-p53 OP43T (Millipore); anti-RPA-p34 (Ab-1; Thermo Scientific, Waltham, MA); and anti-Rad51 (3C10 monoclonal antibody [MAb]; Millipore). The primary antibodies used in murine cells were as follows: anti-PML clone 36.1-104 (Millipore) and anti-53BP1 (Bethyl Laboratories, Montgomery, AL). Cy2, Cy3, or Cy5 fluorescence-labeled secondary antibodies, as well as horseradish peroxidase (HRP)-labeled secondary antibodies, were obtained from Jackson Immuno Research, Newmarket, United Kingdom.

SA–β-Gal staining.

Cells grown on glass coverslips were fixed in 4% formaldehyde for 5 min and stained for senescence-associated β-galactosidase (SA–β-Gal) activity as described previously (46). The cells were quantified using an Axioplan 2 microscope with a 63× Plan-Apochromat oil objective (Carl Zeiss, Jena, Germany).

Microscopy and focus quantification.

Images were acquired with a Zeiss LSM 510 laser scanning confocal device attached to an Axioplan 2 microscope using a 63× Plan-Apochromat oil objective (Carl Zeiss, Jena, Germany). For quantification of γH2AX foci and PML NBs, as well as association events, optical sections (∼0.5 μm) of the whole-cell nucleus were taken and reduced to one plane by maximum projection. For quantification with the open source software ImageJ and customized macros, images were processed as described previously (47). Association events were scored with the colocalization macro available on the ImageJ website (http://rsbweb.nih.gov/ij/) and were defined as the overlap of PML and γH2AX signals in at least 5 pixels.

SIM.

Structured illumination microscopy (SIM) was performed on an Elyra S.1 system from Carl Zeiss using a C-Apo 63× oil objective with a numerical aperture of 1.4. An Andor iXon 885 EM charge-coupled device (CCD) camera served as the detector. The gain of the camera was set to a maximum of 10 to maintain a wide dynamic range. For multicolor acquisition, the sample was sequentially imaged using grids matched to the different wavelengths. Five rotations were used for maximum homogeneous resolution enhancement in the lateral plane. For three-dimensional (3D) SIM, stacks with a step size of 120 nm were recorded. For reconstruction of SIM images, the SIM-processing tool of the built-in ZEN software was used in automatic mode. Refinement of the reconstruction was done manually by adapting the noise filter. Chromatic aberrations of the optical system were determined and corrected with the built-in channel alignment tool of the ZEN 2011 software using TetraSpeck 0.1-μm beads (Life Technologies) as fiducial markers.

Theoretical association probability.

To calculate the theoretical probability of a colocalization event between sphere-shaped objects (PML bodies and irradiation-induced foci [IRIF]) in the nucleus, the volume of the nucleus and the number and volume of the objects were taken into account. These data were determined from 3D image stacks from immunofluorescence experiments. In the case of a colocalization (or association) event of an IRIF and a PML body, the distance (d) between the centers of the two colocalizing distinct spheres within the nucleus is equal to or less than the sum of the spheres' radii: d < r_g + r_p, where r_g is the radius of the genomic locus and r_p is the radius of a PML body. Considering a single nucleus of radius R with m PML bodies of radius r_p each and n IRIF with radius r_g each, we assume that the positions of all IRIF are equally distributed within the nucleus and that no two IRIF are closer than 2 r_g + 2 r_p, which guarantees that a specific PML body does not colocalize with more than one IRIF at a time. Then, considering the volume of the nucleus and the total volume of the spheres around all IRIF, which define a colocalization event, the colocalization probability, P, for one PML body is given by the following equation: P = n (r_g + r_p)³/R³. As for m PML bodies, the probability, P_m, that no PML body colocalizes with any IRIF is given by the following equation: P_m = (1 − P)^m. We get the final result for the probability, P_A, of finding at least one colocalization between a PML body and an IRIF within the whole nucleus with m PML bodies and n genomic loci from the following equation: P_A = 1 − [1 − n (r_g + r_p)³/R³]^m. This model contains the explicit assumption that two IRIF have a minimum distance of 2(r_p + r_g), which is a plausible assumption regarding the sizes of IRIF and of PML bodies. On the other hand, PML bodies are allowed to intersect in this probability model, which does not coincide with our experimental findings. Nevertheless, for the numbers of PML bodies in our experiments, the effect of this approximation on the true probability is very small. From the colocalizing probability, P_A, within one nucleus, one can compute the probability, P_k(i), for the event to find among k investigated cells i cells with at least one colocalization caused by chance as follows: P_k(i) = (k/choose i) P_Aⁱ (1 − P_Aⁱ)^{(k − i)}. If in an experiment with k cells there are q cells with at least one colocalization, then the P value for this event is the sum of the probabilities P_k(i) to find q cells or more with at least one colocalization: P value = SUM(q ≤ i ≤ k) P_k(i).

RESULTS

PML bodies are nonrandomly associated with transient and persistent DNA damage foci.

In order to obtain a quantitative assessment of the association between IRIF and PML bodies, we analyzed their localization in WI-38 fibroblasts at several time points after low (2-Gy)- or high (15-Gy)-dose γ-IR. Irradiation with 2 Gy and 15 Gy induced a dose-dependent DNA damage response detected as increased levels of p53 and p21, but only the 15-Gy treatment led to induction of cellular senescence, as judged by increased levels of p16 and SA–β-Gal (Fig. 1A to C). Interestingly, there was decreased expression of p21 (but not p53) 3 days after 2-Gy irradiation and decreased expression of p53 and p16 (but not p21) 3 days after 15-Gy irradiation (Fig, 1A). We observed these expression patterns in several independent experiments using Wi-38 fibroblasts and therefore exclude experimental artifacts. These results suggest a 2-wave activation pattern for at least some cell cycle-related factors after irradiation. Growth curves and phospho-pRB levels revealed that irradiation with 2 Gy induced a transient cell cycle arrest with proliferation resuming 4 days after irradiation, while 15-Gy-irradiated cells displayed senescence-associated permanent cell cycle arrest (Fig. 1A and D). Irradiation did not induce cell death (Fig. 1E). As expected, increased numbers of IRIF were detected 30 min after irradiation (Fig. 2A and B). While the number of IRIF decreased to control levels after 24 h in 2-Gy-irradiated cells, nuclei of 15-Gy-irradiated cells contained 10 ± 3 persistent DNA damage foci 6 days after irradiation (Fig. 2B).

FIG 1.

Irradiation-induced senescence in WI-38 fibroblasts. (A) WI-38 fibroblasts were gamma irradiated with 2 Gy or 15 Gy, and whole-cell lysates were analyzed by immunoblotting to detect the indicated proteins. d, days. (B) Representative images of SA–β-Gal staining of WI-38 cells 6 days after γ-IR. (C to E) Quantification of SA–β-Gal staining (C), living cells (D), and dead cells (E) within 6 days after irradiation. At least 50 to 100 cells per time point and irradiation dose were monitored. All experiments were done in triplicate. Mean values ± standard errors of the mean (SEM) are depicted.

FIG 2.

PML NBs nonrandomly associate with persistent DNA damage foci. (A) Indirect immunofluorescence staining of WI-38 fibroblasts after γ-IR. Unirradiated (0 Gy) and irradiated cells (2 Gy or 15 Gy) were fixed at the indicated time points and stained with antibodies against PML and γH2AX. Scale bar, 5 μm. The inset in the bottom right image is an enlargement of the boxed area. (B to F) Quantification of γH2AX foci (B), association events between PML NBs and γH2AX (C), percentage of γH2AX foci associated with PML NBs (D), PML NBs (E), and theoretical probability of one random association between PML NBs and γH2AX foci (F). Quantifications of foci and potential associations were carried out using ImageJ software and customized macros. Mean values ± SEM are depicted.

Coimmunostaining of endogenous PML protein revealed that only a very small number of IRIF are associated with PML bodies during the first 24 h after 2-Gy irradiation (Fig. 2A and C). At later time points, the number of IRIF-PML body associations was as low as in control cells (Fig. 2C). Association between PML bodies and IRIF was pronounced in 15-Gy-irradiated cells (Fig. 2A, inset). The number of IRIF-PML body associations was high at all time points after irradiation with 15 Gy (Fig. 2C). The association rate peaks 1 day after irradiation (6 ± 1) and declines to 3.5 ± 1.2 after 6 days. Qualitatively, we observed only a few association events per nucleus irrespective of the time point of analysis and the damage load. Quantitation of association events per nucleus revealed that not more than 30% of IRIF were associated with a PML body at all time points and irrespective of the damage load (Fig. 2D).

Since we also observed an irradiation-induced increase in the number of PML bodies (Fig. 2E), the possibility that an IRIF/PML body association occurs randomly in the nucleus could not be fully excluded. We therefore employed a self-developed algorithm (48) to calculate the theoretical probability of association events between IRIF and PML bodies. This required determination of the number and volume of PML bodies and IRIF and the nuclear volume at several time points after irradiation (data not shown). With these data, we determined that the theoretical probability of only one random association between IRIF and PML bodies is extremely low (Fig. 2F). Thus, the topological association between PML bodies and IRIF is highly nonrandom at all time points for up to 1 week during a gamma irradiation-induced DNA damage response in human fibroblasts. Because the frequency of PML body-IRIF contacts in nonirradiated fibroblasts was at least 10-fold higher than the theoretical value for such contacts, we conclude that PML bodies also assemble at persistent (irreparable) DNA damage foci in fibroblasts without experimental genotoxic stress.

To further analyze the spatial relationship between PML bodies and transient or persistent DNA damage foci, we applied different doses of a UVA laser microbeam to induce DNA damage at 2 or 3 distinct loci in single nuclei. After irradiation with 2 μJ per pulse, Wi-38 cells displayed DNA damage foci, all of which disappeared 24 h after irradiation, most likely after successful DNA damage repair (Fig. 3A to D and data not shown). These transient DNA damage foci were found to be associated with PML bodies 3 h after irradiation, corroborating the notion that PML bodies can specifically associate with transient DNA damage foci (Fig. 3C, yellow arrows). One day after beam irradiation with 2 μJ per pulse, the γH2AX signals had disappeared almost completely. Only application of extreme artificial contrast stretching revealed residual γH2AX dots, which did not colocalize with PML bodies (Fig. 3D, enlargements with asterisks). Microbeam irradiation with 4 μJ or 8 μJ per pulse produced DNA damage foci associated with PML bodies as early as 30 min after irradiation (Fig. 3E and I, yellow arrows). This indicates that the timing of PML body assembly at IRIF is a function of the amount of damage within an IRIF. A UVA dose of 4 μJ or more induced persistent IRIF, which were not resolved after 24 h in Wi-38 fibroblasts. These unrepairable IRIF were associated with one or more PML bodies at all time points of observation (Fig. 3E to L).

FIG 3.

PML NBs associate with UVA microirradiation-induced foci. WI-38 fibroblasts were irradiated with different doses of a UVA microbeam, fixed at the indicated time points, and immunostained to detect PML (green) and γH2AX (red). The regions within nuclei marked by white boxes are shown as magnified views (separated into single color channels shown as grayscale images) on the right of each overview image. (D) Note that in the images marked by asterisks, the signals for γH2AX were contrast stretched to visualize residual fluorescence. Scale bars = 5 μm.

Next, we analyzed the spatial relationship between PML bodies and persistent DNA damage foci in replicative senescent human and mouse primary fibroblasts. As expected, Wi-38 fibroblasts showed upregulation of the senescence markers p16, p21, p53, and SA–β-Gal during serial passaging (Fig. 4A). Sixteen percent (±4%) of γH2AX foci in proliferating fibroblasts (population doubling [PD] 42) were associated with PML bodies (Fig. 4B, yellow arrow, and H). Replicative senescent Wi-38 fibroblasts at PD 58 accumulated 14 ± 3 persistent SDF per nucleus, 35% of which were found to be associated with a PML body (Fig. 4C and H). Freshly isolated wild-type (wt) MEFs contained very low numbers of γH2AX foci (Fig. 4D). Senescence was induced using gamma irradiation or a combination of the DNA-damaging drugs bromodeoxyuridine and distamycin A (BD) or by serial passaging (Fig. 4E to G). These treatments induced a 4- to 5-fold upregulation of the number of DNA damage foci, and 30% to 40% of these foci were associated with a PML body (Fig. 4H). These data show that PML bodies specifically assemble at DNA damage foci in senescent primary human and mouse fibroblasts. However, the majority of these foci are not in contact with a PML body (Fig. 4C, E to G, red arrowheads, and H). In order to assess if PML body-damage focus contacts are random or specific, we compared the number of actually observed contacts with the theoretical probability of random associations determined with a computer program (48) (see Materials and Methods) (Table 1). This analysis showed, for example, that the probability of only one random contact between a PML body and a persistent DNA damage site is ∼7%. This means that in 7 out of 100 cells, one contact can occur randomly. However, we measured on average 5 such contacts in every cell nucleus. Similarly, the probability of the randomness of the observed association rates in irradiated cells at different time points after irradiation was always below 1%, indicating a specificity of interaction at least 2 orders of magnitude higher than random association events (Table 1).

FIG 4.

PML NBs associate with DNA damage foci in senescent human and murine cells. (A) WI-38 fibroblasts at different PDs were subjected to Western blot analysis using antibodies against the indicated proteins (top). The number of SA–β-Gal-positive cells during serial passaging of these cells was also monitored (bottom). α-tub, α-tubulin. (B to G) Representative midnucleus confocal images of young WI-38 fibroblasts (B), replicative senescent WI-38 fibroblasts (C), unirradiated primary MEFs (pMEFs) (D), 15-Gy-irradiated MEFs after 6 days (E), MEFs treated with BrdU and DMA for 6 days (F), and replicative senescent MEFs (passage 6; PD 12) (G). The cells were fixed and immunostained to detect PML (green) and γH2AX (red). The percentages of of SA–β-Gal-positive cells are indicated (means ± SD). (H) Quantification of γH2AX foci (gray bars) and percentages of γH2AX foci associated with PML NBs (blue bars) for the indicated cells. Freshly isolated primary MEFs (passage 1) were used for irradiation and drug treatment, and replicative senescent MEFs were at passage 6 (PD 12). Experiments were done in triplicate. Mean values ± SEM are depicted.

TABLE 1.

Association between PML bodies and γH2AX foci is highly nonrandom^a

Status	Time after irradiation	No. of γH2AX foci	No. of PML NBs	No. of associations per nucleus	Probability of observed association rate being random (%)	Probability of 1 random association (%)
2-Gy irradiation	0	5	10	1	0.14	0.14
	0.5 h	29	8	2	0.01	1.42
	1 day	6	10	1	0.93	0.93
	3 days	6	11	1	0.82	0.82
	6 days	4	13	1	0.22	0.22
15-Gy irradiation	0	5	10	1	0.14	0.14
	0.5 h	62	10	5	0	1.09
	1 day	27	13	6	0	3.92
	3 days	8	14	3	0	1.12
	6 days	10	24	3	0	5.43
Replicative senescence		13	25	5	0	7.28

^aThe average number (n = 100 nuclei each) of γH2AX foci and of PML bodies and their association rate were determined in young Wi-38 fibroblasts after irradiation and in replicative senescent cells as indicated. The probabilities of these associations were determined using a self-developed computer algorithm (48) (see Materials and Methods).

Previously, it was reported that irradiation-induced DNA damage foci in mouse lung cells are associated with PML bodies (19). We sought to extend these observations by analyzing nonirradiated tissue. Immunofluorescence analyses of thin sections of human skin samples revealed a high number of PML body-phosphorylated ATM (pATM) focus associations (Fig. 5A and B). PML bodies were found to be associated with 30% ± 18% and 47% ± 12% of pATM-containing DNA damage foci in cells of the epidermis and dermis, respectively (Fig. 5C). These observations demonstrate that a close physical relationship between PML bodies and DNA damage foci does not occur only after irradiation but may also be a hallmark of unperturbed human tissue. PML body-DNA damage focus association was also observed in human skin tissue when anti-γH2AX antibodies were used (Fig. 5D). pATM foci in cell nuclei of tissues are indicative of persistent DNA damage foci in senescent cells (16). Senescent cells in human skin tissue are characterized by the accumulation of annexin A5 (ANX5) in the periphery of the cell nucleus (45). When ANX5 and pATM were immunostained simultaneously in epidermal skin layers, we observed that the majority of pATM focus-positive cells also expressed ANX5, indicating that they were senescent cells (Fig. 5E and F). This provides strong support for our conclusion that PML bodies are associated with persistent DNA damage foci in human skin. Taken together, our data demonstrate a highly nonrandom association between PML bodies and IRIF or senescence-associated DNA damage foci in various cell culture systems, as well as in normal human skin.

FIG 5.

PML NBs associate with DNA damage foci in human skin cells. Cryosections of human skin were stained with antibodies against PML (green) and phosphorylated ATM (red). (A and B) Representative confocal images of the epidermis (A) and the dermis (B). The insets are enlargements of the boxed areas. (C) The percentage of pATM-positive cells, as well as the association between pATM foci and PML NBs, was quantified in five skin sections from different donors. Mean values and SEM are depicted. (D) Antibodies against γH2AX (red) also revealed contacts between DNA damage sites and PML bodies (green) in the dermis of human skin. (E) Confocal image of a human skin section immunostained with antibodies against ANX5 (green) and pATM (red). DNA was stained with DAPI. (F) Quantitation of the cells positive for the indicated events from experiments shown in panel E. Mean values and SEM are depicted (n = 10 individual healthy skin sections from 3 different middle-aged donors).

DNA damage response factors do not accumulate in PML bodies during senescence.

PML bodies attached to SDF may contribute to sustained DNA damage signaling in senescent cells. In order to address this hypothesis, we analyzed whether DDR factors become transiently associated with PML bodies during a DNA damage response. Control experiments confirmed that in Wi-38 fibroblasts all PML bodies also contained Sp100 during a DNA damage response, in DNA damage-induced senescent cells, and in replicative senescent fibroblasts (data not shown). Therefore, anti-Sp100 antibody staining is an appropriate means to detect PML bodies in coimmunostaining analyses. MDC1 accumulated at γH2AX foci 30 min after irradiation with 2 Gy (Fig. 6B). One day after irradiation, γH2AX foci had disappeared, and MDC1 resumed its diffuse nucleoplasmic distribution (Fig. 6C). Importantly, at no time point after irradiation with 2 Gy or 15 Gy did we observe any colocalization of MDC1 with PML bodies (Fig. 6B to E, insets). Likewise, neither 53BP1 nor phosphorylated NBS1 was found to colocalize with PML bodies adjacent to unrepaired DNA damage foci (Fig. 6D to G). PML bodies appeared to be excluded from γH2AX-labeled chromatin, while the DDR factors completely colocalized with the γH2AX signal (Fig. 6D to G, insets). The same result was obtained in replicative and DNA damage-induced senescent fibroblasts (Fig. 6H and I). In irradiation-induced senescent Wi-38 fibroblasts, p16, p21, and p53 were found diffusely distributed in a micropunctate pattern throughout the nucleus (Fig. 6J to L). Although some of the microdots spatially coincided with the position of a PML body in confocal images (Fig. 6J to L, line scans), there was no substantial enrichment of p16, p21, or p53 within the centers of PML bodies (Fig. 6J to L, insets). Our data, therefore, indicate that the DDR factors MDC1, 53BP1, phospho-NBS1, p16, p21, and p53 are not components of PML bodies during a DNA damage response or in senescent cells.

FIG 6.

MDC1, 53BP1, pNBS1, p16, p21, and p53 do not accumulate at persistent DNA damage-associated PML NBs. WI-38 fibroblasts were gamma irradiated and fixed at the indicated time points after DNA damage induction (A to H) or passaged until senescence (I). The fixed cells were subjected to immunofluorescence staining. An antibody against Sp100 was used as a marker for PML NBs, combined with a γH2AX antibody to stain DNA damage foci. Additionally, the DNA damage response proteins Mdc1, 53BP1, and phosphorylated Nbs1 were stained. (J to L) WI-38 fibroblasts were gamma irradiated with 15 Gy, fixed after 6 days, stained with antibodies against p16, p21, or p53, and costained with an antibody against PML. Insets show enlargements of the boxed area in each image. The line scans (right) show the pixel intensity distribution along the lines indicated in the merged image. Scale bars = 5 μm.

During DNA repair, all PML bodies are in contact with RPA/Rad51-containing repair foci.

To analyze the association between PML bodies and sites of DNA damage during ongoing repair in more detail, we applied 3D SIM. SIM provides an optical resolution limit of ∼100 nm. Human fibroblasts were irradiated with 10 Gy and immunostained 3 h later to detect PML bodies, γH2AX foci, and RPA (Fig. 7A). At that time point, cell nuclei contained 25 ± 7 PML bodies and 68 ± 17 γH2AX foci (mean ± standard deviation [SD]; n = 20 nuclei). The superresolution images revealed that 91% ± 7% of all γH2AX foci in a cell nucleus (n = 10) overlapped with at least one RPA focus (Fig. 7A-1 and -2). Twenty-two percent (±5%) of the latter were also in contact with a PML body (Fig. 7A-1). Interestingly, γH2AX foci without an associated RPA focus were never found in contact with a PML body, while all RPA foci without an associated γH2AX focus were always in contact with a PML body. In other words, during ongoing DNA repair (3 h after 10-Gy irradiation), all PML bodies were in contact with either a γH2AX or an RPA focus or with both at the same time. Very similar observations were made when Rad51 foci were analyzed in combination with γH2AX foci and PML bodies by SIM (Fig. 4J).

FIG 7.

Superresolution imaging of PML nuclear bodies and DNA damage sites. Primary human fibroblasts (at low PD) were irradiated with 10 Gy, fixed after 90 min, and subsequently immunostained to detect the indicated proteins. 3D SIM superresolution images of single nuclei were collected using an Elyra structured-illumination microscope (Zeiss). The white boxes within the overview images are shown enlarged on the right. Each image on the right also shows the individual channels in monochrome.

PML depletion does not substantially impair the DNA damage response.

As PML NBs associate with DNA damage foci, the question arises as to whether cells lacking PML show an impaired DNA damage response. To investigate this, we used primary MEFs freshly isolated from PML wild-type, heterozygous, or knockout embryos, as well as primary human foreskin fibroblasts stably expressing a short hairpin RNA (shRNA) targeting all nuclear PML isoforms. These cells did not express detectable amounts of PML protein or PML nuclear bodies (reference 48 and data not shown). Western blot analyses of irradiated MEFs revealed no differences in the irradiation-induced phosphorylation of H2AX, KAP1, or 53BP1 when cells with or without endogenous PML were compared (Fig. 8A). Measurement of the protein levels from several independent Western blots also showed no statistically significant differences in the activation levels of these DDR factors in PML-proficient versus -deficient MEFs after irradiation (Fig. 8B and C). We then analyzed DNA damage focus formation in PML knockout MEFs. Irradiation with 2 Gy induced formation of numerous γH2AX foci after 30 min, and these foci completely colocalized with 53BP1 immunostaining (Fig. 9A). DNA damage focus formation in MEFs was quantified over time after irradiation with 2 or 15 Gy. The results indicated no significant difference in the efficiency of focus formation (Fig. 9B and C). These observations show that PML does not play an important role in IRIF formation and the DNA damage response in primary MEFs. In contrast, PML-depleted human fibroblasts showed a significantly reduced number of DNA damage foci 30 min and 3 h after 2-Gy irradiation but not at later time points (Fig. 9D). However, IRIF formation was not changed in PML-depleted human fibroblasts after 15-Gy irradiation (Fig. 9E). These observations suggest a functional role for PML in IRIF maintenance during the DSB repair process after low-dose but not high-dose irradiation in human fibroblasts.

FIG 8.

DNA damage signaling is not altered in PML knockout MEFs. (A) PML wild-type, heterozygous, and knockout MEFs were gamma irradiated and lysed at the indicated time points, and Western blot analyses were carried out using the indicated antibodies. (B and C) Relative protein levels as measured by densitometry of Western blot bands were quantified using ImageJ software. All experiments were done in triplicate. Mean values ± SEM are depicted. Note that the blots shown in panel A have been spliced from two Western blots each. The splice sites are indicated by a thin white line between 3 and 8 h or 0.5 and 3 h within the left and the right blots, respectively. All Western blots shown in panel A originate from the same gel run and were developed simultaneously under the same conditions, which also allows comparison of the relative band intensities even between the 2-Gy- and 15-Gy-irradiated cells.

FIG 9.

IRIF formation in PML-depleted cells. (A) Freshly isolated MEFs from PML wild-type, heterozygous, and knockout embryos were irradiated with 2 Gy and immunolabeled 30 min later using antibodies against γH2AX (green) and 53BP1 (green). The images show representative midnucleus confocal sections. Bars = 5 μm. (B and C) MEFs as in panel A were irradiated with 2 Gy or 15 Gy, and samples on coverslips were taken at different time points after irradiation. The cells were immunolabeled as in panel A to quantify the DNA damage foci. The data points represent mean values ± SEM. (D and E) Same as in panels B and C, using primary human fibroblasts stably expressing a vector control, a vector expressing control siRNA, and a vector expressing siRNA against PML. The data points represent mean values ± SEM. ***, P ≤ 0.001 (t test statistical analysis comparing data points for siControl and siPML).

Lack of PML does not alter DNA damage-induced or replicative senescence.

Next, we addressed the question of whether DNA damage-induced senescence is impaired in the absence of PML. Control and PML-depleted human fibroblasts were irradiated with 15 Gy. As shown by SA–β-Gal staining, the lack of PML has no significant effect on the efficiency of senescence induction after 6 days (Fig. 10A). A similar result was obtained when BrdU/DMA was used to induce senescence (Fig. 10B). BrdU/DMA treatment also induced persistent DNA damage foci in freshly isolated wild-type MEFs, but their number was not altered in the absence of PML (data not shown). Notably, the efficiency of BrdU/DMA-induced senescence is not changed in PML^−/− MEFs (Fig. 10C). Similarly, there is no difference between wt and PML^−/− MEFs after radiation-induced senescence (Fig. 10C). These data demonstrate that the PML protein is dispensable for DNA damage-induced senescence in primary human and murine fibroblasts. Finally, we analyzed the replicative potential of human and mouse fibroblasts expressing or lacking PML. Primary human fibroblasts ceased proliferation after 90 to 100 days of serial passaging (Fig. 10D). Surprisingly, PML-depleted fibroblasts reached PD 38, while PML-containing cells yielded about 10 more population doublings. This result suggests that a stable knockdown of PML in primary human fibroblasts acts as an antiproliferative. We observed this phenomenon in several independent stable human PML knockdown cell lines (data not shown). In contrast, MEFs entered replicative senescence-induced proliferation arrest after 10 to 12 days irrespective of the endogenous PML expression level (Fig, 10E). These observations suggest that PML does not play a major functional role in the course of replicative senescence, at least in MEFs.

FIG 10.

PML depletion does not alter fibroblast senescence. (A and B) Primary human fibroblasts with a stable PML knockdown (siPML) or control-infected cells (vector and siControl) were gamma irradiated with 15 Gy and fixed after 6 days (A) or treated with 50 μM BrdU and 10 μM DMA and fixed at the indicated time points (B). To monitor senescence induction, cells were stained for SA–β-Gal. (C) pMEFs isolated from PML wild-type, heterozygous, and knockout embryos were treated with 50 μM BrdU and 10 μM DMA or 20-Gy irradiation, fixed after 6 days, and stained for SA–β-Gal activity. In all experiments, at least 50 cells were monitored, and all experiments were done in triplicate. Mean values ± SEM are depicted. (D and E) HFFs with a stable PML knockdown (siPML) or control-infected cells (vector and siControl) (D), as well as wild-type (+/+), heterozygous PML knockout (+/−), and PML knockout (−/−) pMEFs (E), were cultured until they reached senescence, and population doublings over time were monitored. For pMEFs, 2 or 3 cell lines per genotype were used, and mean values ± SEM are depicted.

DISCUSSION

Numerous models exist for the role of the PML protein or the nuclear bodies in the DDR. It was suggested that NBs may serve as DNA repair compartments (27), as DNA damage sensors (28, 30), or as a modification platform for DNA damage-activated p53 and other DDR factors (23, 39, 42, 43, 49, 50). Telomeric and nontelomeric DNA damage efficiently initiates and maintains senescence (9). Since PML has been functionally implicated in senescence induction (21, 22, 33, 40), we expected to find a functional link between PML and DNA repair and DNA damage-induced senescence. Surprisingly, our data demonstrate that this is not the case for human and murine primary fibroblasts.

PML bodies associate with persistent DNA damage foci in vitro and in vivo.

We confirmed that PML NBs nonrandomly associate with persistent DNA damage foci after γ-IR, UVA irradiation, or radiomimetic-drug exposure (17, 19, 30). In extension of such analyses, we observed that an association between PML bodies and IRIF occurs only at late time points after damage induction by γ-IR or after high-dose and spatially concentrated UVA irradiation. In contrast, low doses of γ-IR (≤2 Gy), which inflict repairable DNA damage in primary fibroblasts (Fig. 1A and B), did not increase the association rate between PML bodies and IRIF. We therefore conclude that PML bodies do not act as immediate-early sensors of DNA damage. This assumption is corroborated by our observation that 2 μJ UVA microbeam irradiation induces PML body-IRIF associations only 3 h after the irradiation pulse (Fig. 3). On the other hand, higher doses of UVA induced PML-IRIF associations as early as 30 min, consistent with previous observations (30). The timing of PML recruitment may therefore be a function of the number of DSBs in an IRIF (28).

Notably, only about 30% or 50% of persistent DNA damage foci are associated with a PML body in vitro or in vivo, respectively. In HCA2 human foreskin fibroblasts, the frequency of association appears to be higher (19). Nevertheless, although the number of contacts may differ between cell types, nuclei always contain DNA damage foci that are not in contact with a PML body. Dellaire and colleagues could show, with the help of electron microscopy, that the structure of these persistent foci does not depend on the spatial proximity to a PML NB (16). Thus, PML bodies are dispensable for the maintenance of persistent DNA damage foci.

Persistent DNA damage has been proposed to be an important trigger for senescence induction (4, 6, 51, 52). This suggested that an association of PML with persistent foci is a marker for presenescent and senescent cells (19). Here, we confirmed this assumption for human skin. These foci, also referred to as SDF (4, 9, 53), may represent clustered DNA damage sites (54, 55), difficult-to-repair DNA damage in heterochromatin (56), or uncapped telomeres (4, 5, 57). DNA synthesis was not detected in persistent foci (19). However, as they still recruit DNA damage factors, it is very likely that they transmit a DNA damage signal, which is sufficient to maintain a DDR and also to establish cell cycle arrest and senescence (51, 52). This is consistent with the observation that self-organized assembly of canonical DDR factors on chromatin in the absence of DNA damage is sufficient to induce a DDR and cell cycle arrest (51). Since DNA damage-induced senescence is not compromised in fibroblasts lacking PML (Fig. 10), we conclude that PML nuclear bodies are dispensable for senescence-maintaining DDR signaling from IRIF or persistent foci. Although the PML body-associated pATM foci observed here in human skin tissue fulfilled all criteria that define persistent DNA damage foci (size, morhopology, ANX5-positive senescent cells, and association with PML bodies) (11, 13) (Fig. 5E and F), we cannot fully exclude the possibility that these pATM foci are of a transient nature.

Potential function of PML at IRIF.

The present study confirmed that PML bodies associate with irradiation-induced, as well as endogenous, persistent DNA damage foci (17, 19). These observations suggested a model in which PML bodies form at sites of persistent DNA damage, from where DNA damage response signaling may be regulated (17, 18). The model implied that focus formation may be impaired in the absence of PML. Analyses in the present report revealed that the formation of DNA damage foci is not impaired in PML-deficient human or mouse primary fibroblasts after senescence-inducing doses of gamma irradiation. However, human primary fibroblasts lacking PML show significantly reduced numbers of IRIF after low-dose irradiation (2 Gy) (Fig. 9D). This is consistent with recently published studies showing that accumulation of many DDR factors at DNA damage foci is abolished in PML-depleted cancer cells (18, 58–60). In primary human embryo lung fibroblasts, the clearance of γH2AX foci after 10 Gy of irradiation was shown to be reduced in PML-depleted cells (60). Interestingly, the formation of IRIF was not impaired in this setting, since the number of γH2AX foci after 30 min of irradiation was not changed (60). In contrast, here at 2 Gy we observed a significantly reduced number of IRIF in PML-depleted human fibroblasts after 30 min of irradiation (Fig. 9D). Collectively, these observations suggest that PML (bodies) may contribute to both formation and clearance of IRIF. In contrast to this conclusion, depletion of PML in HT 1885 fibrosarcoma cells has no impact on γH2AX levels and formation of Rad51 foci (59). The discrepancy from our and previous (60) results may be explained by differences in the experimental setup, mainly the use of a cancer cell line in the study by Yeung et al. (59). Immortal cells may have compromised DNA damage response activities compared to the primary cells used in our study. Furthermore, we consider our analysis by counting individual IRIF at the single-cell level more sensitive than determination of γH2AX fluorescence over a population of cells. In contrast to human cells, IRIF formation after 2-Gy irradiation was not impaired in PML knockout MEFs (Fig. 9B), indicating cell- or species-specific roles for PML in DDR.

Homology-directed repair (HR) on artificial reporter plasmids is impaired in PML-depleted BJ/tert and HT1885 cells (58, 59). PML bodies normally do not contain DNA (16, 61) but accumulate factors involved in DNA repair, especially during HR, e.g., BLM (26, 62–64), WRN (60, 65), RPA (19, 26, 30, 66), and BRCA1 (18). As this pathway involves extensive formation of single-stranded DNA (67), PML bodies may help to organize single-stranded chromatin structure during HR. This is consistent with the observations (i) that PML bodies are associated with ssDNA foci, (ii) that formation of these foci is impaired after PML depletion, and (iii) that PML KO MEFs have an elevated rate of sister chromatin exchange (27, 64). We have shown here for the first time by superresolution microscopy that during ongoing irradiation-induced repair virtually all PML bodies are engaged at γH2AX, Rad 51, or RPA foci, or combinations of them (Fig. 7). It had been shown previously for human fibroblasts that both Rad51 and RPA specifically accumulate at irradiation-induced ssDNA-containing nuclear foci (68). Together, these observations strongly imply a function for PML nuclear bodies in the formation and/or maintenance of ssDNA-containing repair foci. Since γH2AX foci without an associated RPA or Rad51 focus were never found in contact with a PML body, we suggest that PML bodies assemble at γH2AX foci only when these contain ssDNA.

The superresolution images also revealed that PML bodies, γH2AX foci, and RPA or Rad51 foci overlapped with each other but never fully colocalized. This is in contrast to published data showing accumulation at PML bodies of many DDR factors (18, 19, 26, 30, 62–66). The discrepancy may be explained by the use of different cell types in the various studies. We believe that our observations are more conclusive because we used superresolution microscopy on primary cell types. In addition, many previous studies used cancer cell lines, including alternative lengthening of telomeres (ALT)-positive lines. ALT is a DNA recombination event that occurs at specialized PML bodies (APBs) at telomeres with telomeric DNA inside the APBs. Finally, some of the previous studies used a prepermeabilization step before fixation of cells. It has been shown recently that such protocols can induce the collapse of adjacent cellular complexes, including nuclear substructures, potentially leading to fixation-induced colocalization events (69).

PML bodies are nonessential in DNA damage-induced cellular senescence.

There is strong evidence that PML is involved in cellular senescence based on overexpression of PML isoform IV (PML 3) (32, 34, 35, 70, 71). PML knockout or knockdown fibroblasts appear to be resistant to Ras-induced and transforming growth factor beta (TGF-β)-induced senescence (32, 40, 42, 72). In the present report, we investigated the role of PML in DNA damage-induced senescence and found that, in contrast to Ras- or TGF-β-induced senescence, the function of PML is not essential for senescence induction after DNA damage in primary human and murine fibroblasts (Fig. 10). In contrast, thymocytes from PML^−/− mice are partially resistant to irradiation-induced apoptosis (37), again suggesting cell-type-specific functions of PML.

We also could not detect an accumulation of the cell cycle inhibitors p53, p21, and p16 at endogenous expression levels at PML bodies (Fig. 6). This was surprising, as other groups have observed stress-induced colocalization of PML and p53 (17, 32, 42, 50, 73). Finally, at least PML-depleted primary MEFs are not resistant to or show accelerated replicative senescence (Fig. 10E). In contrast, primary human fibroblasts lacking PML displayed somewhat reduced replicative potential, as they did not reach the maximum number of population doublings observed in control fibroblasts (Fig. 10D). This was unexpected, since PML is believed to have tumor suppressor activities. We have no ready explanation for this phenomenon, and further studies are required to unravel the mechanisms of the stable PML depletion-induced proliferation inhibition during long-term passaging of human fibroblasts.

In conclusion, our data show that the association of PML NBs with persistent DNA damage foci is characteristic of senescent cells in cell culture and in vivo. In contrast to Ras- and TGF-β-induced senescence, the function of PML NBs is not essential for senescence induction after DNA damage in cells from both species. However, we stress that PML association at persistent DNA damage sites should still be considered to be a cellular signal to promote apoptosis and senescence, and PML probably is just one of several partly redundant factors that may provide this signal. Our findings extend and refine our understanding of the role of PML NBs in tumor suppression. It is becoming apparent that more data are needed on the effect of different stressors and senescence inducers in analyses of PML functions. Also, careful discrimination between the signal pathways in different cell types and species is necessary when PML is investigated with respect to stress-induced cellular responses.