Continuous elimination of oxidized nucleotides is necessary to prevent rapid onset of cellular senescence

Abstract

Reactive oxygen species (ROS) appear to play a role in limiting both cellular and organismic lifespan. However, because of their pleiotropic effects, it has been difficult to ascribe a specific role to ROS in initiating the process of cellular senescence. We have studied the effects of oxidative DNA damage on cell proliferation, believing that such damage is of central importance to triggering senescence. To do so, we devised a strategy to decouple levels of 8-oxoguanine, a major oxidative DNA lesion, from ROS levels. Suppression of MTH1 expression, which hydrolyzes 8-oxo-dGTP, was accompanied by increased total cellular 8-oxoguanine levels and caused early-passage primary and telomerase-immortalized human skin fibroblasts to rapidly undergo senescence, doing so without altering cellular ROS levels. This senescent phenotype recapitulated several salient features of replicative senescence, notably the presence of senescence-associated beta-galactosidase (SA beta-gal) activity, apparently irreparable genomic DNA breaks, and elevation of p21^Cip1, p53, and p16^INK4A tumor suppressor protein levels. Culturing cells under low oxygen tension (3%) largely prevented the shMTH1-dependent senescent phenotype. These results indicate that the nucleotide pool is a critical target of intracellular ROS and that oxidized nucleotides, unless continuously eliminated, can rapidly induce cell senescence through signaling pathways very similar to those activated during replicative senescence.

When propagated in culture, normal somatic cells achieve a limited number of divisions before undergoing the loss of proliferative capacity termed replicative senescence (1). Several studies have suggested that cell senescence plays a role in organismic aging (2, 3) and that activation of senescence programs in cancer cells block tumor progression (4, 5). Consequently, elucidating the biochemical mechanisms of cellular senescence is critical for understanding the physiologic basis of aging and the mechanisms of tumorigenesis.

Several lines of evidence indicate that cumulative damage to cellular constituents sustained during culture in vitro eventually triggers senescence (6, 7). Such damage can be inflicted by reactive oxygen species (ROS), which are byproducts of incomplete mitochondrial electron transfer (8). Despite the actions of detoxifying enzymes, such as superoxide dismutases (SOD1 and SOD2) and catalase, and low molecular weight antioxidants, increasing oxidative stress due to age-related mitochondrial dysfunction may eventually exceed the capacity of cellular antioxidant defenses (9). Indeed, both ROS levels and oxidative damage levels are found to be higher in late-passage cells relative to early-passage cells (10). Additionally, increased oxidative stress in the form of hyperoxia (11), treatment with hydrogen peroxide (12), shRNA-mediated knockdown of SOD1 (13), or oncogenic Ras overexpression (14, 15), causes cells to enter senescence prematurely. Conversely, culturing cells at low ambient oxygen can significantly delay the onset of senescence (16), supporting a causal role for ROS in triggering several forms of cell senescence.

Attempts at defining the precise role(s) of ROS in senescence and possible cellular targets of ROS action have been confounded by the multiple biochemical and cell-physiologic actions of these radicals. ROS are not only able to damage biomolecules but can also act as intracellular second messengers (15, 17). Hence, the ability of ROS to induce senescence may reflect either the inflicting of cellular damage or physiologic signaling or a combination of both.

The induction of the DNA damage-checkpoint response pathway (18, 19), ostensibly caused by the persistence of unrepaired double-strand DNA breaks (DSBs) (20), is a common feature of several different forms of senescence. However, the precise origins of such DSBs and the contribution of elevated ROS levels to the accumulation of DNA damage have remained unclear.

To distinguish the effects of oxidative DNA damage from the multiple effects of elevated cellular ROS levels, we developed an experimental model in which endogenous oxidative DNA damage could be increased without a concomitant elevation of ROS levels. To do so, we manipulated the levels of the most frequently occurring oxidized DNA lesion, 8-oxoguanine (8-oxoG), which is found at elevated levels not only in senescent cells (21–23) but also in the frontal brain cortex (24) of elderly individuals. Deoxyguanosines (dG) may suffer oxidation at two stages–either when present in the soluble nucleotide pool or after incorporation into DNA. Once formed, the oxidized nucleotide, 8-oxo-dGTP, is readily incorporated into DNA by DNA polymerases.

Recent evidence suggests that the detoxification of 8-oxo-dGTP by the major cellular 8-oxo-dGTPase, MTH1 (Mammalian MutT Homolog 1) (25) represents the dominant mechanism responsible for maintaining low 8-oxodG levels in human DNA (26). Accordingly, to increase total cellular levels of 8-oxoguanine, we used small hairpin interfering RNAs (shRNA) to suppress MTH1 protein expression. Our studies show that knockdown of MTH1 rapidly induces the senescence program in early-passage human skin fibroblasts and in hTERT-immortalized human fibroblasts and can limit the proliferation of Ras-transformed tumorigenic cells as well.

Results

Induction of Senescence in Primary Fibroblasts After Knockdown of MTH1.

We undertook to assess the effects of suppressing MTH1 levels in young [population doubling (PD) 34–50] human BJ fibroblasts by infecting them with one of two distinct puromycin-selectable lentiviral vector constructs expressing shRNAs directed against MTH1. Stably infected cells were selected by addition of puromycin to the culture media. Both constructs suppressed MTH1 expression, one by ≈78% and the other by >90%, as assessed by Western immunoblotting (Fig. 1A), and both induced senescence within one population doubling, as determined by senescence-associated beta-galactosidase (SA beta-gal) activity (Fig. 1B) and proliferation curves (Fig. 1C). The lack of proliferation was not due to cell death, as measured by propidium iodide and Annexin V staining, which did not vary significantly between the control and shMTH1-infected cells (Fig. S1A). MTH1 suppression also induced senescence very rapidly in MRC5 human fetal lung fibroblasts (Fig. S2 A–C).

Fig. 1.

Introduction of shMTH1 rapidly induces senescence in young fibroblasts. Young BJ skin fibroblast cells (PD 34) were infected with the indicated lentiviral plko.puro constructs and were thereafter cultured in puromycin-containing media starting 24–36 h after infection. (A) Cells were subjected to immunoblotting, using an MTH1-specific antibody to assess the relative abilities of the two MTH1-specific shRNAs to knockdown MTH1 protein. The empty plko viral vector and shRNA against GFP were used as negative controls. Actin was used as a loading control. (B) Cells were stained for SA beta-gal activity 3 days (≈1 PD) after infection with the indicated lentiviral constructs. The percentage of positively stained cells is indicated to the right of the respective images. (C) Population doubling (PD) rates. The indicated cells were plated at an initial density of 4 × 10⁵ and allowed to grow for the time indicated before counting. Cells were replated at the same density at each point on the curve. Note that the shMTH1–1 infected cells begin redividing, because of inefficient MTH1 suppression. (D) Total cellular 8-oxoG levels. FITC-avidin staining is used to detect total cellular 8-oxoG. DAPI staining indicates nuclear DNA. To minimize background, the image contrast was changed to an equivalent extent for all samples. Representative fields are shown. (E) Immunoblotting to detect up-regulation of tumor suppressor proteins involved in senescence induction. Cells were harvested 4–5 days after lentiviral infection and lysed. One hundred micrograms of protein was run on a gel and probed with antibodies against human p53, p21, and p16^INK4a.

Total cellular ROS levels were measured in the control shGFP-infected and in the shMTH1-infected cells. We found that the shRNA-mediated suppression of MTH1 did not substantially affect cellular ROS levels within the first 4–6 days subsequent to infection, a period when most of the experiments described below were carried out (Fig. S1B). There was a noticeable increase in the ROS levels of shMTH1–2 infected cells ≈11 days after infection, consistent with the idea that the senescent state itself leads to elevated ROS levels (27). However, onset of the senescence phenotype clearly preceded this relatively minor increase in ROS levels. This conclusion was reinforced by our observation that neither the levels of etheno-adducts, which serve as markers of acute oxidative stress (28), nor protein levels of the major cellular antioxidants, catalase and SOD1 (29) were altered in the shMTH1-infected cells relative to control cells (Fig. S1 C and D). Hence, the senescent phenotype created by loss of MTH1 could not be attributed to overall changes in cellular ROS levels.

Additionally, suppression of MTH1 did not lead to any significant alterations in total mitochondrial mass or to disruption of mitochondrial membrane potential (Fig. S3). Collectively, these observations rule out a role for mitochondrial aberrations in mediating the shMTH1-induced senescent phenotype, although we cannot rule out that mitochondrial mutations may eventually accrue as a delayed consequence of MTH1 suppression.

Because the outgrowth of cells with incomplete MTH1 suppression is strongly selected, cultures infected with the less efficient shRNA vector, shMTH1–1, were overgrown after 3–4 days of drug selection by cells that had resumed active proliferation, doing so by losing MTH1 inhibition (Fig. 1C and Fig. S7B). However, cultures infected with the more efficient shMTH1–2 showed no proliferation whatsoever for more than a week (Fig. 1C). For this reason, we used the shMTH1–2 construct alone in the majority of experiments described below.

To assess the functional effects of MTH1 suppression, total cellular 8-oxoG levels, including the soluble nucleotide, nucleoside, and free base pools and any incorporated products, were gauged by immunofluorescent FITC-Avidin staining, using a protocol that minimized diffusion of the labile products (30). To test for the specificity of staining, we preblocked the FITC-Avidin solution with soluble 8-oxoguanosine, which was indeed able to reduce the intensity of staining (Fig. S4A). We found that, as anticipated, increased staining intensity was observed upon MTH1 knockdown relative to shGFP infection in both BJ and MRC5 cells (Fig. 1D, Fig. S2E). We also immunostained late-passage (PD 90) BJ human fibroblasts and found that these cells also exhibited the expected elevated levels of total cellular 8-oxoG relative to their early passage (PD 32) counterparts (Fig. S4A). Hence, with respect to 8-oxoG accumulation, loss of MTH1 expression appeared to mimic the effects on BJ cells of extended propagation in vitro.

To ensure that the accumulation of cellular 8-oxoG was not simply a consequence of entrance into a state of cell senescence, cells treated with doxorubicin, a topoisomerase II poison that produces DNA strand breaks, were stained with FITC-Avidin. These cells also underwent senescence, as indicated by the increased beta-galactosidase staining, but did not exhibit up-regulated 8-oxoG levels (Fig. S4B).

We also found that BJ cells entering senescence after shMTH1 knockdown up-regulated the levels of the p53, p21^Cip1/Waf1 and p16^INK4a tumor suppressor proteins; expression of these proteins was rapidly induced within 4–5 days of initial infection with the shRNA constructs, consistent with the observed rapid induction of senescence (Fig. 1E). This was consistent with the notion that the tumor suppressor pathways that mediate replicative senescence (31) also regulate shMTH1-induced senescence. We also observed increased p53 and p21 protein levels in the shMTH1-infected MRC5 cells. However, p16^INK4a levels in these cells were already high even in the control shGFP-infected cells and did not change materially upon introduction of shMTH1–2. (Fig. S2D)

The observed activation of the p53/p21 pathway, which is associated with genotoxic stress, suggests that loss of MTH1 might be inducing senescence through incorporation of 8-oxo-dGTP into high molecular weight DNA rather than simply through its accumulation in the soluble nucleotide pool. To resolve between these alternatives, we treated cells with aphidicolin, a DNA synthesis inhibitor, after lentiviral infection with the control and shMTH1 shRNAs. The SA beta-gal assay demonstrated that the aphidicolin treatment prevented shMTH1-induced senescence (Fig. S5 A and B), despite an equivalent level of MTH1 suppression in the both treated and untreated cells (Fig. S5C). Consequently, we concluded that the main stimulus to the observed entrance into senescence depended on DNA replication, either by incorporation of the already-oxidized nucleotides into DNA or by some other effect of these nucleotides on replication fork progression, rather than simply through accumulation of the free oxidized nucleotide within cells.

Molecular Pathways Involved in Induction of shMTH1-Induced Senescence.

Introduction of hTERT, the catalytic subunit of human telomerase, can overcome replicative senescence in BJ human fibroblasts (32). Because suppression of MTH1 appears to induce a senescent phenotype with characteristics that closely resemble those of replicative senescence, we undertook to determine whether BJ cells immortalized by the expression of hTERT would also be resistant to shMTH1-induced senescence. We found, however, that hTERT-expressing human BJ fibroblasts (BJ-hTERT) also underwent senescence within 1–2 PDs of MTH1 knockdown, as gauged by population doubling rates and the presence of acidic beta-galactosidase activity (Fig. 2A). Hence, the repeatedly observed induction of senescence did not derive from changes in the genome that could readily be reversed by the actions of telomerase.

Fig. 2.

Molecular pathways mediating shMTH1-induced senescence. (A) Introduction of hTERT does not overcome shMTH1-induced senescence. BJ cells containing the hTERT vector (denoted as BJ-T) were infected with shMTH1–2 or shGFP. A population doubling curve was established and the cells were fixed and stained for SA beta-gal activity 4 days after infection with the constructs (Inset). The percentage of positively stained cells in indicated below the representative images. (B) Introduction of SV40 large T antigen prevents the shMTH1-induced proliferation defect. BJ cell lines immortalized with hTERT and SV40 Large T antigen (denoted as BJE-LT) were infected with shMTH1–2 and shGFP. Three days after infection, 2 × 10⁵ cells were plated and counted in triplicate every subsequent day. (C) Abrogation of the p53 pathway is sufficient to rescue shMTH1-induced senescence. To block the p53 pathway, the neomycin-selectable construct pLXSN. HPV-E6 was introduced into BJ PD37 cells, and the resulting cells were cultured in neomycin (G418)-containing media for 10 days. These cells were subsequently infected with shGFP or shMTH1–2, and population doublings were assessed. (D) Indicated sample lysates (75 μg) were immunoblotted with the anti-MTH1 antibody to ensure equal knockdown in the E6-expressing and control cells and with p21 and p53 antibodies. BJ PD89 cells are shown as a positive senescent control. (E) SA beta-gal staining. The control and HPV.E6-expressing BJ cells were fixed and stained for SA beta-gal activity. The percentage of positive cells is indicated below a representative field for each sample.

In response to MTH1 knockdown, these BJ-hTERT cells also exhibited increased p21/p53 and p16 protein levels (Fig. S6A). In contrast, BJ-hTERT cells in which both the p53/p21 and p16/Rb pathways were functionally inactivated through expression of the SV40 large T antigen (denoted BJE-LT cells) did not show the shMTH1-induced proliferation defect (Fig. 2B and Fig. S6 B and C). This observation reinforces the idea that either one or both of the major tumor suppressor pathways that mediate replicative senescence also regulate shMTH1-induced senescence.

To determine whether one tumor suppressor pathway was more important than the other in mediating the MTH1 suppression-induced senescent phenotype, BJ cells were infected with a vector expressing the human papilloma virus type E6 (HPV.E6) oncoprotein, an antagonist of the p53 pathway (33). These cells were then infected with either shGFP or shMTH1–2 and their tendencies to undergo senescence were assessed. Although the doubling rate of the BJ.E6 shMTH1-infected cells was slightly decreased relative to that of the BJ.E6 shGFP-infected cells, the cells did not undergo a proliferative arrest (Fig. 2C). As anticipated, when compared with control cells infected with shMTH1, the BJ.E6 shMTH1 cells showed no up-regulation of either p21 or p53 protein levels (Fig. 2D). Moreover, only a minority of the BJ.E6 shMTH1-infected cells developed SA beta-gal activity (Fig. 2E).

In contrast, introduction of HPV.E7, which abrogates Rb/p16 tumor suppressor signaling (33), was unable to prevent the shMTH1-dependent senescent phenotype to the same extent (Fig. S6 D and E). Therefore, despite the up-regulation of p16 protein levels in BJ cells, the p53 pathway appears to be the dominant tumor suppressor pathway involved in the induction of shMTH1-induced senescence, similar to what is observed for replicative senescence in BJ cells (34). The absence of p16 protein up-regulation in the shMTH1-infected MRC5 fibroblasts, relative to their shGFP-infected counterparts, also suggests a minor role, if any, of the p16^INK4a pathway in mediating MTH1 suppression-induced senescence (Fig. S2D).

Suppression of MTH1 Expression Induces Genomic DNA Damage.

Genomic DNA damage, and particularly unrepaired DNA double-strand breaks (DSBs), are known to be features of replicative senescence (18). We reasoned that the presence of clustered 8-oxoG lesions in chromosomal DNA could lead to reduced efficiency of the base excision repair process leading, in turn, to persistent DNA single -strand break repair intermediates (35); the latter could then act as DSB-prone sites in subsequent rounds of DNA replication.

To determine whether suppression of MTH1 resulted in DNA strand breaks, we used the single-cell gel electrophoresis procedure, often termed the “comet assay” (36), which involves lysing agar-embedded cells in situ and subjecting them to electrophoresis under alkaline conditions. The alkaline conditions convert any abasic sites (indicative of incomplete base excision repair) into single-strand breaks. The size of any resulting “comet” tail corresponds to the extent of damaged DNA released from the nuclei and can be visually categorized as none, medium, or intense (Fig. 3A). As we observed, the percentage of cells with intense comet tails nearly quadrupled for the young BJ cells infected with the shMTH1 constructs relative to the shGFP control. This analysis revealed the creation of single- and double-strand breaks in chromosomal DNA after suppression of MTH1 (Fig. 3A); these breaks survived for an extended period in an unrepaired state subsequent to their formation. Hence, failure to detoxify 8-oxo-dGTP results in significant chromosomal DNA damage, which derives ostensibly from the incorporation of this oxidized base into the DNA.

Fig. 3.

Genomic DNA damage in shMTH1-infected cells. (A) Single cell gel electrophoresis (comet) assay. Three days after infection, the lentiviral-infected samples and BJ PD86 (as a positive control) were subjected to alkaline unwinding and electrophoretic conditions as described in Materials and Methods. Examples of the three different types of cells scored visually for each sample are shown with the comet tail indicated by arrows. (B) Genomic 8-oxo-dG quantitation. Genomic DNA was isolated from indicated cells and subjected to LC-MS/MS analysis to quantitate 8-oxo-dG/10⁶ nucleotides. Fold-changes were normalized with respect to absolute numbers of 8-oxodG/10⁶ nucleotides from the untreated samples, and the values are indicated in the table below the respective samples. Except for shGFP (which has an n = 1 because of technical problems with the samples in experiments 2 and 3), each value represents an average of 3 experiments (corresponding raw data are in Table S1). (C) Double strand break (DSB) foci formation. DAPI staining indicates nuclear DNA. Colocalization of the red γH2AX and green 53BP1 foci is indicated by the presence of yellow foci in the merged fields. Contrast was equivalently enhanced for all samples to improve image quality. Cells with 3 or more bright punctate foci in each field were scored as positive, with 6–8 fields being counted per sample (in duplicate), totaling over a hundred cells. Percentage of foci-positive cells is indicated below a representative field for each sample.

To determine whether loss of MTH1 activity led directly to incorporation of the oxidized nucleotide into DNA, we gauged levels of genomic 8-oxo-dG by high performance liquid chromatography and triple quadrupole mass spectrometry (LC-MS/MS) (28). We found that the shMTH1 cells consistently showed a small (from ≈1.4- to 2-fold) steady-state increase in 8-oxo-dG relative to control untreated cells, despite baseline fluctuations of genomic 8-oxo-dG between samples (Fig. 3B and Table S1). Such low levels are not surprising, given that these measurements are made in the background of ongoing base excision repair activity, which can rapidly remove 8-oxo-dG from genomic DNA. The presence of significant DNA breaks/abasic sites observed under alkaline comet assay conditions is indicative of initiation, but not completion, of such repair (Fig. 3A). Nevertheless, despite the stochastic variations in base excision repair activity and the unavoidable minor fluctuations in culture conditions, both of which can affect genomic 8-oxo-dG measurements, our measurements repeatedly indicate a trend of increased incorporation of 8-oxo-dGTP into genomic DNA when MTH1 expression is suppressed.

Because irreparable double-strand breaks (DSBs) are thought to be the major inducers of senescent phenotypes (19), we undertook to detect the presence of DNA DSBs by costaining cells that had been infected with either the shGFP or shMTH1–2 vectors with antibodies against the gamma-H2AX (γH2AX) histone variant and 53BP1, both of which are known to bind sites of chromosomal DSBs and are involved in signaling the presence of such breaks via the ATM/Chk2 pathway (18). MTH1 knockdown by infection with the shMTH1–2 vector led to a greater fraction (2.6- to 4-fold higher) of total cells with 3 or more well-defined and colocalized foci of gamma-H2AX/53BP1 relative to cells infected with the control shGFP vector (Fig. 3C). Although the observed foci are different in appearance from the numerous, discrete foci normally observed when cells are treated with chemical DSB-inducing agents, they appear identical in relative number and appearance to the DSB foci seen in BJ cells that have entered replicative senescence after extended passage in vitro (ref. 20, data not shown). Previous work has shown that the threshold of DNA damage required to initiate cellular senescence can be as low as one or two DSBs per cell (37). These observations therefore provided strong indication that increased oxidized guanines in the soluble nucleotide pool are sufficient to generate the genomic DNA damage and persistent DSB signaling that are characteristic of replicative senescence.

Reduced shMTH1-Induced Senescence in Cells Cultured in 3% Oxygen.

Low ambient oxygen tensions are associated with reduced oxidative stress (38, 39) and reduced G-to-T transversions, these being the mutations that are the signatures of genomic 8-oxo-dG (40). We reasoned, therefore, that culturing BJ cells under conditions of low oxygen tension should reduce endogenous 8-oxo-dGTP levels, thereby mitigating the need for MTH1 expression. Accordingly, age-matched BJ cells were cultured in parallel in the presence of either 21% oxygen or 3% oxygen. As expected, the cells grown at 3% oxygen had lower total ROS levels relative to the cells cultured at 21% oxygen (Fig. S7A). Both sets of cells were then infected with vectors expressing either shGFP or shMTH1–2, resulting in equal levels of MTH1 knockdown in both sets of cells (Fig. 4C). Under conditions of MTH1 knockdown, the populations cultured in 21% oxygen exhibited an ≈3-fold greater frequency of SA beta-gal-positive cells than did those cultured in 3% oxygen (Fig. 4A). Indeed, the cells grown at 3% oxygen were able to proliferate after suppression of MTH1 expression, in contrast to those cultured at 21% oxygen, which failed to do so (Fig. 4B).

Fig. 4.

Reduced shMTH1-induced senescence in cells cultured in 3% oxygen. BJ PD34 cells were cultured at either 21% (normox) or 3% (hypox) oxygen tension before and after infection with either shGFP or shMTH1–2. (A) SA beta-gal staining. Cells were fixed and stained 4 days after infection with the indicated lentiviral constructs. Percentage of SA beta-gal cells is indicated below the respective images. (B) Population doubling curves for BJ PD34 (at 21% and 3% oxygen) infected with either the shGFP or shMTH1–2 vector. (C) Immunoblotting was carried out on the specified samples (70 μg) using antibodies against MTH1, p21 and p53. Actin was used as the loading control. (D) Detection of total cellular 8-oxoG, using FITC-avidin. DAPI staining indicates nuclear DNA.

The shMTH1-infected cells grown in low oxygen also exhibited lower p53 and p21 protein levels relative to their counterparts cultured in 21% oxygen (Fig. 4C) and lower total cellular 8-oxoG levels (Fig. 4D), reinforcing the observation that the signals triggering senescence in response to MTH1 knockdown are significantly reduced in low-oxygen culture. Furthermore, the shMTH1-infected cells could be cultured at 3% oxygen for an additional 3 weeks (≈10 PDs) while maintaining MTH1 knockdown (Fig. S7B). In contrast, shMTH1-infected cultures grown at 21% oxygen were dominated by the outgrowth of MTH1-expressing cells after only a week or so (Fig. S7B), indicating the strong selection that favors cells expressing normal levels of MTH1. This observation underscores the strong influence of ambient oxygen in determining the rate with which 8-oxo-dGTP is generated in cultured cells and the resulting reduced requirement for MTH1 function in cells grown under conditions of low oxidative stress.

Discussion

We have shown that increased concentrations of oxidized purine nucleotides in the soluble nucleotide pool, resulting from suppression of the MTH1 8-oxo-dGTPase, are sufficient to induce senescence in early-passage human fibroblasts and in their hTERT-immortalized counterparts. This senescent phenotype is induced without any measurable increase in intracellular ROS levels. The features of MTH1 loss-induced senescence resemble replicative senescence; both exhibit an activation of the p53 and p16 tumor suppressor pathways, increased genomic DNA damage, and the induction of DSB repair-associated proteins. Together, these observations indicate that replicative senescence can occur as a direct response to oxidative DNA damage, much of it affecting guanine nucleotides. Indeed, the rapid initiation of the senescence program in early-passage cells upon MTH1 knockdown implies that even robustly proliferating cells are poised perilously close to the brink of senescence, and that they are protected from entering this state by efficient antioxidant defenses including, notably, MTH1.

Although senescent cells are known to accumulate irreparable DNA double-strand breaks (18, 20), it has not been clear why they do so. Given the heightened vulnerability of the cytosolic nucleotide pool to oxidation (41) when compared with chromatin-bound genomic DNA, it seems plausible that oxidized DNA precursors, generated continuously over the course of a cellular lifespan and surviving because of incomplete detoxification, find their way into genomic DNA and, with some frequency, create DSB-prone sites. The present results suggest that these DSBs result from incorporation of oxidized nucleotides followed ostensibly by incomplete genomic repair.

The phenomenon of increased oxidized dNTP incorporation can be exacerbated by deterioration of the DNA damage response. Thus, age-related declines of both cellular DSB repair (42) and base excision repair (BER) (43) have been reported. Inefficient BER, in particular, can convert a lesion such as an 8-oxoG base into a DNA strand break, because the repair pathway initially creates a nick at the site of lesion removal, followed by a rate-limiting step that involves insertion of the correct base and religation of the DNA backbone. We have detected persistent nuclear Apurinic/apyrimidinic Endonuclease 1 (APE1) foci in MTH1-suppressed cells, similar in appearance and number to the DSB foci, which suggest that incomplete or dysfunctional BER may be the source of the observed DSBs (P.R. and R.A.W., unpublished data).

Several studies have assessed changes in genomic 8-oxo-dG levels upon senescence and have found that they are increased to various extents (12, 21, 22). Our quantitative results, despite statistical fluctuations, nevertheless show a trend of increase in steady-state genomic 8-oxo-dG levels upon loss of MTH1. Importantly, our measurements of the relative increases in genomic 8-oxo-dG levels are very likely to be underestimates, because the true extent of shMTH1-dependent changes initially inflicted on genomic 8-oxo-dG is likely to be obscured by ongoing repair activity present in most cells. In support of our results, other studies have correlated suppression of MTH1 with reduced processing of 8-oxo-dGTP (41) and, conversely, increased MTH1 protein levels with lower genomic 8-oxo-dG (26). Independent of the detailed mechanisms, the present observations described here reveal a hitherto unexplored means by which oxidative stress and DNA damage, long observed in older cells and organisms, can directly affect proliferative lifespan.

Materials and Methods

Detailed methods are available in SI Methods.

Plasmids and Constructs.

The MTH1 over-expression construct was subcloned (as described in SI Methods) from a plasmid expressing full-length WT human MTH1 (pcDEB.MTH1), a gift from Dr. Yusaku Nakabeppu at Kyushu University. The pBp.HPV-E7 construct was a gift from Dr. Karl Munger's laboratory at Harvard Medical School.

Antibodies and Western Blot Analysis.

Immunoblotting procedures were performed as described in ref. 44.

Viral Production and Infection of Target Cells.

Lenti- and retroviral production and infection of target cells were as described in ref. 45. Cells were selected with 2 μg/ml puromycin or 250 μg/ml G418 (neomycin).

Cell Culture.

BJ, BJ-T, BJ-LT and MRC5 cell lines were maintained in DME medium supplemented with 10% FCS (D10 media), 100 units/ml penicillin, 100 μg/ml streptomycin and 2 mM l-glutamine at 37 °C in either 21% oxygen/5% CO₂ or, where specified, in 3% oxygen/5% CO₂.

Proliferation Rate Measurements.

To determine the average rate of population doubling (PD), 4 × 10⁵ cells were plated in duplicate and the number of cells was counted every 3 days, with 4 × 10⁵ cells being replated for the next count. The numbers were converted into population doublings, using the following formula: [log (no. of cells counted) − log (no. of cell plated)]/log (2). To assess cell division rates, 1 × 10⁵ or 2 × 10⁵ cells were plated in a set of 12 plates for a 4-day growth curve and on each subsequent day, plates were counted in triplicate.

Senescence-Associated Beta-Galactosidase (SA beta-gal) Assay.

Detection of SA beta-gal activity was carried out as described in ref. 46. For quantification purposes, a minimum of 100 cells, spanning 5–6 different microscopy fields, were scored for staining.

8-oxoG Detection and DSB Foci Analyses by Immunofluorescence.

Total cellular 8-oxoG was analyzed as described in refs. 30 and 47.

For detection of DNA DSB foci, cells were fixed in 4% paraformaldehyde (16% stock, Electron Microscopy Sciences), permeabilized in 0.1% PBST, and blocked in 0.1% PBST/10% FBS. Antibodies were diluted in blocking buffer and incubated at 4 °C overnight.

Quantitation of Genomic 8-oxo-dG.

DNA was isolated from cells using a genomic DNA isolation kit (Roche). Etheno-dA, etheno-dC and 8-oxo-dG were quantified by LC-MS/MS as described in ref. 28.

shRNA Expression Vectors.

The shRNA design, lentivirus production and infection was done as described in ref. 45. The target sequences are common to all known transcript variants of MTH1. Target sequences for the MTH1 shRNA constructs are as follows: shMTH1–1: 5′-GTGGCTGCTGAACAGCTGCAA-3′; and shMTH1–2: 5′-GAAATTCCACGGGTACTTCAA-3′. The control shRNA was targeted against GFP.

Single-Cell Gel Electrophoresis (Comet) Assay.

The comet assay was carried out using the CometAssay kit (Trevigen) instructions.