• Open Access

DNA damage response and cellular senescence in tissues of aging mice


  • Chunfang Wang,

    1. Ageing Research Laboratories, Institute for Ageing and Health and Center for Integrated Systems Biology of Ageing and Nutrition (CISBAN), Newcastle University, Newcastle upon Tyne NE4 6BE, UK
    Search for more papers by this author
  • Diana Jurk,

    1. Ageing Research Laboratories, Institute for Ageing and Health and Center for Integrated Systems Biology of Ageing and Nutrition (CISBAN), Newcastle University, Newcastle upon Tyne NE4 6BE, UK
    Search for more papers by this author
  • Mandy Maddick,

    1. Ageing Research Laboratories, Institute for Ageing and Health and Center for Integrated Systems Biology of Ageing and Nutrition (CISBAN), Newcastle University, Newcastle upon Tyne NE4 6BE, UK
    Search for more papers by this author
  • Glyn Nelson,

    1. Ageing Research Laboratories, Institute for Ageing and Health and Center for Integrated Systems Biology of Ageing and Nutrition (CISBAN), Newcastle University, Newcastle upon Tyne NE4 6BE, UK
    Search for more papers by this author
  • Carmen Martin-Ruiz,

    1. Ageing Research Laboratories, Institute for Ageing and Health and Center for Integrated Systems Biology of Ageing and Nutrition (CISBAN), Newcastle University, Newcastle upon Tyne NE4 6BE, UK
    Search for more papers by this author
  • Thomas Von Zglinicki

    1. Ageing Research Laboratories, Institute for Ageing and Health and Center for Integrated Systems Biology of Ageing and Nutrition (CISBAN), Newcastle University, Newcastle upon Tyne NE4 6BE, UK
    Search for more papers by this author

Professor Thomas von Zglinicki, Ageing Research Laboratories, Institute for Ageing and Health, Campus for Ageing and Vitality, Newcastle University, Newcastle upon Tyne NE4 6BE, UK. Tel.: +44 191 248 1104; fax: +44 191 248 1101; e-mail: t.vonzglinicki@ncl.ac.uk


The impact of cellular senescence onto aging of organisms is not fully clear, not at least because of the scarcity of reliable data on the mere frequency of senescent cells in aging tissues. Activation of a DNA damage response including formation of DNA damage foci containing activated H2A.X (γ-H2A.X) at either uncapped telomeres or persistent DNA strand breaks is the major trigger of cell senescence. Therefore, γ-H2A.X immunohistochemistry (IHC) was established by us as a reliable quantitative indicator of senescence in fibroblasts in vitro and in hepatocytes in vivo and the age dependency of DNA damage foci accumulation in ten organs of C57Bl6 mice was analysed over an age range from 12 to 42 months. There were significant increases with age in the frequency of foci-containing cells in lung, spleen, dermis, liver and gut epithelium. In liver, foci-positive cells were preferentially found in the centrilobular area, which is exposed to higher levels of oxidative stress. Foci formation in the intestine was restricted to the crypts. It was not associated with either apoptosis or hyperproliferation. That telomeres shortened with age in both crypt and villus enterocytes, but telomeres in the crypt epithelium were longer than those in villi at all ages were confirmed by us. Still, there was no more than random co-localization between γ-H2A.X foci and telomeres even in crypts from very old mice, indicating that senescence in the crypt enterocytes is telomere independent. The results suggest that stress-dependent cell senescence could play a causal role for aging of mice.


Replicative senescence has been first described by Hayflick (1965) as the irreversible loss of division capacity of primary human cells in vitro after a reproducible number of population doublings (PD). In addition to permanent growth arrest, preferentially in G0, this phenotype is associated with major changes in gene expression patterns (Shelton et al., 1999), morphology and function of cells. Senescence can be induced in primary as well as transformed cells by a variety of stimuli: replicative senescence is the result of telomere uncapping, normally triggered by continuous loss of telomere sequence during cell division (Bodnar et al., 1998). All stressors that result in DNA double strand breaks can induce a senescent phenotype, as can genetic overactivation of the cell cycle check point machinery, overexpression of oncogenes (Lin et al., 1998; Dimri et al., 2002; Ferbeyre et al., 2002), and various other stresses (Campisi, 2003).

DNA double strand breaks as well as telomere uncapping (d’Adda di Fagagna et al., 2003; Herbig et al., 2004; Takai et al., 2004) induce a DNA damage response. This response is characterized by the activation of Ataxia Telangiectasia Mutated (ATM) and RAD-3 related kinase (ATR) (Rouse & Jackson, 2002) being recruited to the site of damage and leading to phosphorylation of Ser-139 of histone H2A.X molecules (γ-H2A.X) adjacent to the site of DNA damage. The phosphorylation of histone H2AX facilitates the focal assembly of checkpoint and DNA repair factors including 53BP1, MDC1/NFBD1 and NBS1, and also promotes the activation by phosphorylation of the transducer kinases Chk1 and Chk2, which converge the signal on p53/p21 and (possibly in human but not in mouse fibroblasts) p16 (Smogorzewska & de Lange, 2002).

Whether senescence is simply an artefact of cell culture or whether it exists to any significant amount in vivo has been repeatedly and intensively debated (Ben-Porath & Weinberg, 2005). Part of the reason for the uncertainty regarding the relevance of cell senescence in vivo is the absence of good markers. The expression of hundreds, if not thousands of genes is significantly changed in senescent cells. However, few of these changes are senescence specific and of sufficient magnitude to be considered as marker candidates. Thus, the best available marker until recently was a histochemical assay for β-galactosidase activity at pH6 (senescence-associated β-galactosidase or Sen-β-gal; Dimri et al., 1995), which was not causally related to senescence induction (Lee et al., 2006), was cumbersome at least in some tissues and resulted not too seldom in contradictory data (Kurz et al., 2000; Severino et al., 2000). We recently proposed to use the formation of DNA damage foci, measured by simple IHC or immunofluorescence (IF), as marker for telomere- or DNA damage-induced cellular senescence in vivo (von Zglinicki et al., 2005). Obviously, senescence is only one possible outcome of a DNA damage response, the other possibilities being DNA repair or apoptosis. However, these two latter outcomes have only a short lifetime and, thus, a low probability of detection, while senescence is a permanent, actively maintained state of DNA damage response (d’Adda di Fagagna et al., 2003), at least as long as growth factor- and stress pathway signalling is not interrupted (Bakkenist & Kastan, 2004; Satyanarayana et al., 2004). Moreover, there are specific assays for apoptosis that allow its separation. In human fibroblasts in vitro under a large variety of conditions, Sen-β-Gal and γ-H2A.X deliver identical results (Passos et al., 2007). There might be cases in which even a continuous DNA damage response does not lead to senescence, for instance in postmitotic cells or during meiotic division. Moreover, some components of the canonical DNA damage response might be expressed at lower levels in ‘deep’ senescence when compared with an immediate response, however, the presence of DNA damage foci correlated well with unrelated markers of cellular senescence in aging mouse liver (Panda et al., 2008).

Telomere dysfunction-induced foci have recently been used to demonstrate the increase with age of senescent fibroblasts in skin of baboons (Herbig et al., 2006; Jeyapalan et al., 2007), animals that are ‘human like’ with respect to telomere length and essential absence of telomerase activity. Laboratory mice have much longer telomeres than long-living mammals and retain high levels of telomerase activity in many tissues during adulthood. Therefore, it is assumed that telomere-dependent replicative senescence does not play a major role in aging of these mice. However, mice fibroblasts in vitro show high levels of oxidative stress (Busuttil et al., 2003) and stress-induced cell senescence (Parrinello et al., 2003). Genetic activation of p53 shortened mice lifespan despite suppression of tumour formation (Tyner et al., 2002; Maier et al., 2004), while a p21 knock-out rescued at least some accelerated aging phenotypes in telomerase (mTERC) knock-out mice (Choudhury et al., 2007), and p16 knock-out rescued various mice stem cell systems from age-related decline (Janzen et al., 2006; Krishnamurthy et al., 2006; Molofsky et al., 2006). Modification of a cellular senescence programme is a likely, but by far not wholly proven, cause of these observations. Thus, we wanted to find out whether cells containing DNA damage foci could be found at all in tissues of mice with long telomeres, and whether the frequency of these potentially senescent cells would increase with aging.

We first confirmed that stress-induced senescence in primary mouse embryonic fibroblasts (MEFs) was accompanied by a significant induction of DNA damage foci. In contrast to those in senescent mTERC−/− fibroblasts, foci in wtMEFs did not significantly co-localize with telomeres. In livers from old mice, we found very similar frequencies of Sen-β-gal- and γ-H2A.X-positive hepatocytes and a good but not perfect spatial correlation between these two markers. We next measured frequencies of foci-positive cells in ten tissues in male C57/Bl6 mice over an age range from 12 to 42 months. Frequencies increased with age in lung, spleen, dermis, liver and intestinal epithelium. We confirmed in skin and gut that these foci did not co-localize with telomeres. Foci formation in the gut was exclusively observed in crypts, while apoptosis was restricted to villi. These data suggest that the frequency of cells in telomere independent, DNA damage-induced senescence increases with age in various tissues of the mouse, consistent with the idea that cellular senescence might be among the possible causes of aging, even in organisms with long telomeres.


DNA damage foci formation in senescent MEFs and in mouse liver

Wild-type MEFs undergo stress-induced cell senescence after 7–10 PD at ambient oxygen concentration in vitro (Parrinello et al., 2003), followed later by spontaneous immortalization because of loss of checkpoint function. We analysed the frequency of DNA damage foci using anti-γ-H2A.X IF in young proliferating, senescent and spontaneously immortalized MEFs and compared it to that in senescent late generation TERC−/− MEFs (Fig. 1A). Quantitative evaluation showed a higher frequency of foci-positive cells in proliferating MEF cultures compared with proliferating human diploid fibroblast (HDF) cultures (Fig. 1B). This is probably due to higher basal levels of oxidative stress in mouse compared with human cells. In fact, growing MEFs under more physiological oxygen tension of 3% significantly reduced the frequency of foci-positive cells from 28 ± 1% to 17 ± 1.8%. Higher oxidative stress in MEFs might also explain the generally higher density and larger size of foci in MEFs compared to HDFs (see Fig. 1A,D). However, frequencies of foci-positive cells in cultures of senescent MEFs were significantly above this background level and as high as in senescent HDFs, irrespective of whether senescence in MEFs was induced by telomere shortening or telomere-independent stress (Fig. 1B). Following spontaneous immortalization, the frequency of foci-positive MEFs returned to background levels (Fig. 1B).

Figure 1.

 DNA damage foci formation in senescent mouse embryonic fibroblasts (MEFs) and mouse liver. (A) γ-H2A.X immunofluorescence (red) in wtMEFS at PD 4.3 (YOUNG), PD 7.4 (senescent, SEN) and PD 9.7 (immortalized, IMMO), as well as in senescent fourth generation TERC−/− MEFs at PD 2.4 (TERC−/−). Nuclei are shown in blue (DAPI). Bar equals 10 μm. (B) Frequencies of foci-positive cells in young, senescent and spontaneously immortalized wtMEFs, senescent TERC−/− MEFs, as well as young and senescent MRC5 human diploid fibroblasts (HDF). Data are mean ± SEM from three independent experiments. Significant differences to young wtMEFs (P < 0.05, TUKEY multiple comparisons) are indicated by asterisks. (C) Correlation between frequencies of Sen-β-Gal-positive and foci-positive MRC5 HDFs grown under various pro- or antioxidative treatments as described (Passos et al., 2007). Data are mean ± SEM from three experiments each. Linear regression and 95% confidence intervals (dotted lines) are given. (D) Immuno-FISH images from senescent wt (top) and fourth generation TERC−/− MEF (bottom). Red: telomeres, green: γ-H2A.X foci, yellow: areas of co-localization. DAPI staining has been omitted for clarity. (E) Average foci-telomere Pearson correlation coefficients for senescent wt and TERC−/− MEFs. Data are mean ± SEM from 100 nuclei each. The difference is significant with P < 0.001 (Student’s t-test). (F) Average frequencies (mean ± SEM, three animals) of hepatocytes positive for Sen-β-Gal (SBG) or γ-H2A.X measured by immunofluorescence (IF), immunohistochemistry on cryosections (IHC) or by immunohistochemistry on paraffin-embedded tissue (IHC/P) in livers from 36-month-old mice. SBG, IF and IHC was performed on adjacent sections from the same animals, whilst IHC/P was measured in three independent mice. IF is different from SBG with P < 0.05. (G) Correlations between SBG and IHC and IF, respectively, in livers from individual 36-month-old mice. Means (dots), regression lines (solid lines) and confidence intervals (dotted lines, for SBG-IHC correlation only) are shown. SBG and IHC are significantly correlated with P = 0.039 (Pearson product moment correlation), but SBG and IF are not. (H) Adjacent cryosections of liver from a 36-month-old mouse stained for Sen-β-Gal (right) or γ-H2A.X (left). Black arrows indicate cells that are positive for both reactions, white arrowheads show cells that are positive for one but not the other reaction.

Next, we subjected HDFs to various pro- or antioxidative treatments (Passos et al., 2007) and measured frequencies of foci-positive cells and of cells positive for Sen-β-gal in parallel. These two indicators of cell senescence correlated with r2 = 0.9882 (Fig. 1C).

Immuno-fluorescence in situ hybridization (FISH) analysis of the co-localization between telomeres and DNA damage foci confirmed the telomere dependency of senescence in TERC−/− MEFs but showed only few, random co-localization spots in senescent wtMEFs despite high foci density (Fig. 1D). Accordingly, the average Pearson correlation coefficient in senescent late generation TERT−/− MEFs was similar to that in senescent HDFs (Passos et al., 2007) but was significantly lower in senescent wtMEFs (Fig. 1E). We conclude in agreement with earlier reports (Sedelnikova et al., 2004) that stress-dependent cellular senescence in MEFs is accompanied and most probably caused by accumulation of nontelomeric DNA damage foci, so that foci formation can constitute a marker for cell senescence in murine cells.

Anti-γ-H2A.X IHC detects a DNA damage response induced by UVB irradiation in mouse skin (Fig. S1, Supporting Information). To see whether anti-γ-H2A.X staining might mark senescent cells in mice tissues, we stained adjacent cryosections from livers of old mice for Sen-β-gal activity and with anti-γ-H2A.X by either IF or IHC. We also prepared paraffin-embedded liver sections from three separate mice following perfusion fixation for γ-H2A.X IHC (IHC/P). Frequencies of γ-H2A.X-positive cells as measured by IHC are very similar to Sen-β-gal-positive cells, while those measured by IF are higher (Fig. 1F). High frequencies of positive cells in mice tissues as measured by IF have also been reported by others (Sedelnikova et al., 2004). There is a significant correlation between IHC and Sen-β-gal results in individual mice (Fig. 1G). The similar slope but higher intercept for the correlation between IF and Sen-β-gal (Fig. 1G) suggests that background might contribute significantly to the high frequencies of positive cells reported by IF. In adjacent sections, at least some of the cells stained by Sen-β-gal are also positive for γ-H2A.X IHC and vice versa (Fig. 1H). We conclude that γ-H2A.X IHC on both cryo- and paraffin sections is a potential biomarker for senescent cells in tissues.

Age dependency of DNA damage foci-positive cells in various mice tissues

We analysed ten different tissues from mice in four age groups between 12 and 42 months by IHC/P. Results are summarized in Table 1. We did not see significant changes in the frequencies of foci-positive cells with age in testis, kidney, eye, heart and skeletal (quadriceps) muscle (Fig. 2 and data not shown). Frequencies of γ-H2A.X- positive nuclei were low in skeletal muscle (Fig. 2A), heart muscle (Fig. 2B) and in kidney epithelial cells (Fig. 2C), while, in seminiferous tubuli in the testis, spermatocytes during meiosis are strongly positive for γ-H2A.X (Fig. 2D), as expected (Lee et al., 2005).

Table 1.   Frequencies of foci-positive cells (in %) in tissues of male C57Bl6 mice at 12 and 42 months of age
Tissues12 months42 months
  1. Data are mean ± SEM, n = 3.

Heart muscle2.8 ± 1.03.0 ± 1.0
Skeletal muscle (quadriceps)0.3 ± 0.51.5 ± 1.2
Kidney2.6 ± 0.81.4 ± 0.7
Eye lens3 ± 2.02.4 ± 1.4
 Seminiferous tubule61.8 ± 10.858.0 ± 12.1
 Hepatocytes8.4 ± 1.717.2 ± 2.6 (P < 0.001)
 Epidermis5.0 ± 3.03.3 ± 1.8
 Dermis1.1 ± 0.24.8 ± 2.5 (P = 0.01)
 Alveoli6.7 ± 0.719 ± 3.6 (P = 0.001)
 Lymphocytes7.3 ± 0.824.9 ± 6.8 (P = 0.003)
Small intestine
 Crypt26.7 ± 1.936.3 ± 2.7 (P < 0.001)
 Villi5.8 ± 2.13.7 ± 1.9
Figure 2.

 Fractions of foci-positive cells in various organs of adult and old mice. γ-H2A.X immunohistochemistry (IHC/P) in skeletal (quadriceps) muscle (A), heart muscle (B), kidney (C), testis (seminiferous tubuli, D), lung (E) and spleen (F) of 12- and 42-month-old mice. Frequencies of positive cells are given for lung (E) and spleen (F). Data are mean ± SEM from three animals per age group. Asterisks mark significant differences (P < 0.05, Student’s t-test).

We found significantly increasing frequencies of foci-positive cells with age in five tissues, namely in lung alveoli (Fig. 2E), white pulp of the spleen (Fig. 2F), crypts of the small intestine, liver hepatocytes and dermis (Table 1).

Age-related losses in barrier functionality, elasticity and thickness (Fig. 3A) of skin have repeatedly been related to cellular senescence in dermis and epidermis (Dimri et al., 1995; Herbig et al., 2006; Jeyapalan et al., 2007). With advancing age, we also find an increasing frequency of foci-positive cells, presumably fibroblasts, in the dermis (Fig. 3B). This increase is significant (Fig. 3C). However, little co-localization between DNA damage foci and telomeres was seen in fibroblasts of the dermis (Fig. 3D) and the average foci-telomere Pearson correlation coefficient was low.

Figure 3.

 Fractions of foci-positive cells increase with age in mouse dermis. (A) Morphology (H&E stain) and (B) γ-H2A.X immunohistochemistry of skin of mice 12 months (left) and 42 months (right) old. (C) Frequencies of γ-H2A.X -positive nuclei (in %) vs. age. Data are mean ± SEM from triplicate staining experiments per animal with three animals per age group. Linear regression (solid line) and 95% confidence intervals (dotted lines) are shown. The slope of the regression is significant with P = 0.0014. (D) ImmunoFISH in the dermis of 36-month-old mice. Marked nuclei are shown at higher magnification in the inserts. Purple: nuclei, green: telomeres, red: γ-H2A.X foci.

In liver, the frequency of γ-H2A.X foci-positive cells increases significantly (P = 0.0037 for trend) with age, confirming and extending published data (Matheu et al., 2007; Panda et al., 2008). It should be noted that the vast majority of foci-positive cells in liver are hepatocytes as judged by morphological criteria (see Fig. 4A, inserts), and only these cells are included in the quantitative evaluation (Fig. 4B). Interestingly, at all ages frequencies of foci-positive hepatocytes in the periportal areas are lower than in the centrilobular and intermediate areas of the lobes (Fig. 4A,B). Foci frequencies in these locations are different with P = 0.0092, and this is independent of age (P = 0.946 for interaction between age and location, two-way anova). To see whether this was correlated to oxidative stress, we stained parallel sections with an antibody against 4-hydroxynonenal (HNE). HNE and other aldehydic lipid peroxidation products are well known to accumulate in liver mitochondria with aging (Chen & Yu, 1994). Most HNE immunoreactivity was found around the central vein, while hardly any of it was seen in the periportal region (Fig. 4C). Differences were highly significant with respect to both location and age (Fig. 4D, P < 0.001 for both, two-way anova).

Figure 4.

 Preferential accumulation of DNA damage foci and 4-hydroxynonenal in the periportal hepatocytes in old livers. (A) Livers of 12-month-old (left) and 42-month-old (right) mice were stained for γ-H2A.X. Examples of periportal and centrilobular areas are shown. pV, portal vein; cV, central vein. Boxed areas are shown at higher magnification below. Foci (in red) are distinct from heterochromatic ‘speckles’. Arrows point to positive nuclei (upper row) or to individual foci (lower row). (B) Frequencies of γ-H2A.X-positive hepatocytes with age in periportal (P) and centrilobular + intermediate (C+I) fields of the liver lobe. Data are mean ± SEM from one staining experiment per animal with three animals per age group. The increase with age is significant and independent of location (P = 0.0037), and the differences between areas are significant (P = 0.0092) and independent of age (two-way anova). (C) Examples of periportal and centrilobular areas of livers stained for 4-hydroxynonenal (HNE). Arrows indicate positive staining in cytoplasm. (D) Frequencies of HNE-positive hepatocytes at locations as above in livers from 12- to 42-month-old mice. Data are mean ± SEM from one staining experiment per animal with three animals per age group, P < 0.001 for both location and age, two-way anova.


A compromised gut function is a common feature of aging in mice. Accordingly, we found foci-positive crypt enterocytes with an increasing frequency at higher age (Fig. 5A). Foci-positive cells were essentially restricted to the crypts (Fig. 5A,B). On the contrary, in the same crypts there were only very few apoptotic cells as defined by DNA degradation detected by apoptag assay. Apoptotic cells were solely confined to the villi (Fig. 5C), which in turn were devoid of γ-H2A.X foci-positive enterocytes (Fig. 5B). Increases in both the frequencies of foci-positive enterocytes in the crypts and apoptag-positive villus enterocytes were significant with age (Fig. 5D,E). Foci-positive cells were clustered in an area slightly above the bottom of the crypt, which is known to contain the fast proliferating, transiently amplifying enterocytes derived from the stem cells that reside close to the crypt base (Marshman et al., 2002). This suggested the possibility that DNA damage foci in the crypts might be caused by replication stress (Ward et al., 2004), and thus be indicative of transiently stalled DNA replication rather than senescence. To test this possibility, we double stained intestinal sections from young and old mice with antibodies against γ-H2A.X and Ki-67 (which recognizes proliferating cells). 80.7 ± 2.5% and 67.7 ± 12.3% of all crypt enterocytes were positive for either the one or the other antibody in 12- and 42-month-old mice, but only 7.3 ± 1.5% and 6.3 ± 3.8% of the enterocytes stained for both (Fig. 5F). This result shows that the contribution of replication stress to DNA damage foci formation in intestinal crypts of aging mice is only a minor one, and that it does not change with age.

Figure 5.

 Differential distribution of senescent and apoptotic enterocytes in aging mice intestine. (A) γ-H2A.X immunohistochemistry of the small intestine of mice 12 (left) and 42 months (right) old. (B) γ-H2A.X immunohistochemistry of the small intestine of a 36-month-old mouse. The boxed area is shown at higher magnification on the bottom. (C) TUNEL staining on the adjacent section to the one shown in (B). (D) Frequencies of foci-positive enterocytes in the crypts vs. age. Data are mean ± SEM from triplicate staining experiments per animal with three animals per age group. Linear regression (solid line) and 95% confidence intervals (dotted lines) are shown. The slope of the regression is significant with P = 0.0017. (E) Frequencies of TUNEL-positive enterocytes in intestinal villi vs. with age. Data are mean ± SEM from three animals per age group. Linear regression (solid line) and 95% confidence intervals (dotted lines) are shown. The slope of the regression is significant with P < 0.0001. (F) Results of a double immunostaining with γ-H2A.X (H2) and Ki-67 (KI). Cells were scored as either single positive (H2+KI− or H2−KI+), double positive (H2+KI+) or double negative (H2−KI−). Data are from three animals per age group with SD generally between 1.5% and 3%.

Oxidative DNA base damage, measured by IHC against 8-oxoG, tended to be higher in the intestine of old animals, although there was a high inter- and intra-organismic variability of the 8-oxoG signal intensity (Fig. S2, Supporting Information).

We did not find a significant association between the frequencies of foci-positive crypt enterocytes and apoptotic enterocytes in the adjacent villi (data not shown). However, villi emanating from crypts staining strongly positive for γ-H2A.X were less well developed than those fed by crypts with less DNA damage foci (Fig. 5B).

To test the impact of telomere uncapping onto senescence in the mouse, we first compared frequencies of foci-positive cells in intestinal crypts in our aging mice to late generation (G4) TERC−/− mice with short telomeres and to their heterozygous littermates, which are similar to wild-type mice in terms of telomere length and lifespan (Blasco et al., 1997). Telomere uncapping in G4TERC−/− mice aged 12–15 months significantly increased the frequency of foci-positive cells to 86 ± 1.4% (mean ± SEM, n = 3), i.e. even beyond that in old wild-type mice consistent with the shortened lifespan of these mice (Rudolph et al., 1999). Interestingly, foci frequencies in crypt enterocytes from heterozygous knock-out mice of around 1 year of age were already significantly higher than in wild-type mice (41 ± 1.8%, n = 5, vs. 26.7 ± 1.9%, n = 3). It is possible that these differences between wild-type and heterozygous mice are due to minor differences in housing conditions or tissue dissection and processing. However, it was recently shown that telomerase can protect mitochondrial function and lower mitochondrial ROS production (Ahmed et al., 2008). An effect of lowered gene dosage of TERC on DNA damage foci formation independent of telomere shortening cannot be excluded, therefore.

Next, we measured telomere length in intestinal enterocytes by quantitative FISH (Q-FISH) (Fig. 6A). Both villus and crypt enterocyte telomere length decreased significantly up to an age of 36 months (Fig. 6B, P < 0.001, Tukey). This confirms and extends published data (Flores et al., 2008). However, no further decrease was seen in 42-month-old animals, which might be due to a survivor effect in these very old animals. In agreement with Flores et al. (2008), telomeres in villus enterocytes were shorter than those in the crypts, and this was significant for all ages (Fig. 6B, P < 0.001, Tukey). However, even in old mice, telomere Q-FISH signals were much stronger than in G4TERC−/− mice enterocytes (Fig. 6A). Together with the absence of DNA damage foci in villus enterocytes this suggested that the amount of telomere loss occurring in the intestine during aging of wild-type mice was not sufficient to cause frequent telomere-dependent senescence. This was confirmed by immuno-FISH analysis: DNA damage foci in crypt enterocytes of wild-type mice did not co-localize with telomeres (Fig. 6C). The spatial association between telomeres and DNA damage foci as measured by the average foci-telomere Pearson correlation coefficient did not change with age and was for all age groups significantly lower than that in late generation TERC−/− crypt enterocytes (Fig. 6D). Together, our data exclude apoptosis, hyperproliferation and telomere-dependent senescence as being significantly associated with DNA damage foci formation in crypt enterocytes, strongly indicating that stress-induced DNA damage causes senescence in this tissue compartment with increasing frequency in aging wild-type mice.

Figure 6.

 Enterocyte telomere shortening with age is not sufficient to cause DNA damage foci formation in the crypts. (A) Telomere Q-FISH in intestinal sections from 12-month-old (left) and 42-month-old mice (right). Upper row: villi, lower row: crypts. For comparison, a representative intestinal section from a G4TERC−/− mouse (age 12 months) is also shown (lower right). (B) Telomere length vs. age in crypt and villus enterocytes from wild-type mice. Data are from 400 to 600 nuclei per compartment and three mice per age group. Box plots indicate median, upper and lower quartiles (boxes), 10th/90th percentiles (whiskers) and 5th/95th percentiles (dots). Differences between crypts and villi (all ages) and the decreases until 36 months are significant (P < 0.001, anova/Tukey). (C) ImmunoFISH on the crypts of a 36-month-old mouse. Marked nuclei are shown at higher magnification at the right. Purple: nuclei, green: telomeres, red: γ-H2A.X foci. Co-localization of foci and telomeres is indicated in yellow. (D) Average foci-telomere Pearson correlation coefficients for crypt enterocytes from wild-type mice at the indicated ages and from G4TERC−/− mice at about 12 months of age. Data are mean ± SD from at least 30 nuclei per mice and three mice per group. Pearson coefficients in all age groups from wt mice are significantly below that in TERC−/− mice, but not significantly different from each other (anova). There is no significant trend with age in wild-type mice (linear regression analysis).


The question whether cellular senescence contributes causally to aging of organs and organisms is obviously of central relevance for any understanding of the aging process. However, so far exhaustive studies of the mere presence of senescent cells in a wide variety of tissues of aging mammals have hardly been performed, with the notable exception of measurements of cell replicative capacity in caloric-restricted mice (Pendergrass et al., 1995; Wolf et al., 1995). Mostly because of technological limitations inherent in any histochemical enzyme activity assay, Sen-β-gal staining, the ‘gold standard’ marker for cellular senescence has most often been used on cells in vitro. Many of the relatively rare in situ studies were devoted to the analysis of cell senescence in individual organs, often in the context of age-related disease rather than aging per se. Examples are vascular aging and atherosclerosis (Fenton et al., 2001; Vasile et al., 2001; Minamino et al., 2002), liver cirrhosis (Wiemann et al., 2002) and osteoarthritis (Martin & Buckwalter, 2002). Analyses of human skin aging by Sen-β-gal staining produced contradictory results (Dimri et al., 1995; Severino et al., 2000). In kidneys of aging rats, senescence of tubular epithelial cells was inferred from increasing positivity for Sen-β-gal, lipofuscin and p16INK4A with age (Melk et al., 2003). More recently, p16INK4A/ARF expression changes with aging were examined in a wide variety of rat and mice organs (Krishnamurthy et al., 2004). p16INK4A mRNA levels increased with age in almost all examined organs. This increase was confirmed at the protein level and attenuated by caloric restriction in some organs. The authors concluded that p16INK4A expression is a biomarker of aging, however, the fact that it increases as well in postmitotic tissues like heart or brain (Melk et al., 2003; Krishnamurthy et al., 2004) casts some doubt on its suitability as a sole marker for cell senescence.

A decreasing percentage of cells free of DNA damage foci with advancing age in five mice tissues (liver, testes, kidney, lung and brain) has been reported before (Sedelnikova et al., 2004). However, the age range covered in this paper was less than the median lifespan. Moreover, the largest changes were found between 2 and 8 months of age, suggesting that most of the observed differences might be related to development rather than aging. Thus, reliable information about the presence of senescent cells during normal aging of mice was still lacking.

Our results indicate a large extent of tissue- and cell type-specificity for cellular senescence in aging mice. Among the essentially postmitotic tissues, we found very little staining in heart and skeletal muscle. Jeyapalan et al. (2007) also found few positive nuclei in baboon skeletal muscle. Whether these might reflect senescence in satellite cells remains open for a more thorough examination.

In the kidney tubular epithelium we detected foci-positive cells; however, their frequency remained low and did not significantly increase between 12 and 42 months of age. This is in apparent contradiction to reported increases in p16 expression and Sen-β-gal activity occurring mostly before 1 year of age (Melk et al., 2003; Krishnamurthy et al., 2004) and in frequencies of γ-H2A.X-positive nuclei measured by IF (Matheu et al., 2007). It is not clear whether this is due to strain-specific or methodological differences.

The increase in foci-positive cells in lung alveoli and spleen corresponds well with a modest increase in p16 expression in these tissues (Zindy et al., 1997; Krishnamurthy et al., 2004). The increase in the frequency of foci-positive dermal fibroblasts with age in mice is significant, but less exaggerated than in the skin of baboons (Herbig et al., 2006; Jeyapalan et al., 2007), possibly because telomere-dependent senescence might be less relevant in wild-type mice with long telomeres.

In mouse liver, Sen-β-gal- and γ-H2A.X staining gave very similar results, suggesting the presence of DNA damage foci as a good marker for hepatocyte senescence. The frequency of foci-positive hepatocytes increased with age similarly to what was shown before in younger age groups (Sedelnikova et al., 2004; Panda et al., 2008) and using p16 as senescence marker (Krishnamurthy et al., 2004). Hepatocyte senescence in aging mice might to a large extent be driven by stress-induced DNA damage because the spatio-temporal distribution of HNE, a good marker for oxidative stress in liver, mirrors closely that of foci-positive cells.

We found a significant increase in γ-H2A.X-positive enterocytes in intestinal crypts with age. We excluded apoptosis and DNA replication stress as major contributors to the observed foci formation and conclude, therefore, that the vast majority of foci-positive enterocytes are in senescence. Telomere shortening to the amount as it occurs with age in wild-type mice intestine apparently does not cause major uncapping and crypt enterocyte senescence, because (i) telomeres in villus enterocytes are significantly shorter than those in crypts but still do not induce DNA damage foci formation, (ii) there is very little co-localization between foci and telomeres in crypt enterocyte nuclei, and this does not increase with age, and (iii) when telomeres become critically short (e.g. in G4TERC−/− mice enterocytes), they induce much higher foci frequencies than those found even in the oldest wild-type mice. We conclude that stress-mediated DNA damage is a major factor to cause cellular senescence in aging mice.

The gut epithelium is constantly renewed. Stem cells reside close to the bottom of crypts and divide slowly. Transiently amplifying enterocytes are fast dividing cells at slightly higher level in the crypts which generate all new enterocytes that move then upwards to replace those cells shed off from the top of the villi. Our data indicate that the transiently amplifying enterocytes in the crypts suffer increasingly from senescence with advancing age. This could seriously compromise the ability of crypts to maintain functional villi. In fact, the areas with strongest γ-H2A.X staining in the crypts did not show any well-developed villi at all (see Fig. 5B). These data suggest crypt cell senescence as a potential cause of gut dysfunction in aging mice.

Taken together, our data show that potentially senescent cells can be detected by conventional IHC on archival, formalin-fixed and paraffin-embedded tissues without the need for fresh tissue samples as required for Sen-β-gal histochemistry (Dimri et al., 1995) or γ-H2A.X IF on touchprints or frozen sections (Sedelnikova et al., 2004; Panda et al., 2008). The use of γ-H2A.X IHC alone might over estimate the frequencies of senescent cells in some organs or cell types, because activated DNA damage response can also lead to apoptosis or repair. Moreover, foci-positive cells in older animals might not show the same activation levels for all components of the DNA damage repair machinery as cells immediately following severe DNA damage (Panda et al., 2008). Clearly, cell senescence is a dynamic process and our knowledge of the defining elements in this process and their kinetics is still limited. However, our data in the intestine show that apoptotic cells and cells in transient repair can be separated out by simple double staining, making γ-H2A.X IHC a convenient and potentially accurate candidate amongst the known markers to measure cellular senescence in vivo.

Experimental procedures

Cell culture

Mouse embryonic fibroblasts from C57Bl6 wild-type, TERC+/− and fourth generation TERC−/− animals were cultured in Dulbecco’s modified Eagle’s medium (Sigma, Dorset, UK) supplemented with 10% heat inactivated foetal calf serum (Sigma), 1%l-glutamine and 1% penicillin/streptomycin. MEFs were grown on 75 cm2 flasks (Corning Incorporated, One Riverfront Plaza, Corning, NY, USA) at 37 °C in an atmosphere of 5% CO2, 20% O2 and 95% air. MEFs were split into fresh medium at 90% confluence.


Three male C57BJ6 mice per age group (12, 22, 36 and 42 months) were killed. Tissues were immediately perfusion fixed. Formalin-fixed paraffin embedded 3 μm sections were prepared from the following tissues: heart, skeletal muscle (quadriceps femoris), testis, kidney, eye, liver, lung, skin, spleen and small intestine. Slides were stored at room temperature until analysed. Livers from three additional old mice (36 months) were snap-frozen and stored at −80 °C. Mice deficient for the template RNA component of telomerase (TERC−/−) were generated in a C57Bl6 background as described (Blasco et al., 1997). Fourth generation (G4) TERC−/− mice showed critically short telomeres and significantly shorter lifespan (Choudhury et al., 2007), while heterozygous TERC−/+ mice were normal with respect to both telomere length and lifespan. TERC−/+ and G4 TERC−/− mice were killed at 12–15 months of age and sections were prepared from small intestine as above.

Histochemistry and immunofluorescence

Paraffin sections were deparaffinized with histoclear and methanol, and antigen was retrieved by incubation in 0.01 m, pH 6.0 citrate buffer at 95 °C for 20 min. Cryosections were air dried for 30 min at room temperature, fixed with 4% paraformaldehyde for 10 min, permeabilized for 45 min with PBG [phosphate-buffered saline (PBS)-containing 0.5% Triton-X 100, 0.5% bovine serum albumin and 0.2% fish skin gelatin] and incubated in 0.3% H2O2 in PBS for 10 min.

Slides were incubated in blocking buffer (Impress anti-rabbit Ig Kit #MP-7401; Vector Lab, Peterborough, UK) for 30–60 min at room temperature. Primary antibodies were applied overnight at 4 °C, then washed three times with TPBS for 5 min. Slides were incubated for 30 min to 1 h with secondary antibody (#MP-7401; Vector Lab). Antibodies were detected using rabbit peroxidase ABC kit (#PK-4001; Vector Lab) according to the manufacturer’s instructions. Substrate was developed using 1% DAB (3,3′-diaminobenzidine tetrahydrochloride; Sigma), 0.3% H2O2 in PBS or with NovaRed (# SK-4800; Vector Lab). Use of NovaRed resulted in higher sensitivity and better signal-to-noise ratios. Therefore, all quantitative data were derived using this substrate. Sections were counterstained with haematoxylin or methyl green. For IF, cryosections were permeabilized for 45 min with PBG and blocked in 1.5% normal goat serum in PBS for 1 h. Incubation with the primary antibody was overnight at 4 °C. Alexa-555 goat anti-rabbit (#A21428; Invitrogen, Eugene, Oregon, USA, 1:6000) was used as secondary antibody for 45 min at room temperature. Tissues were stained with 4′,6-diamidino-2-phenylindole (DAPI) and mounted with DABCo. Cells were regarded as positive if at least one focus per nucleus was visible using a 63× objective in a Leica DM5500B microscope. For quantitative evaluation, 150–200 nuclei were scored per condition.

For IF of MEFs, cells were plated onto coverslips 48 h prior to harvesting. Cells were washed briefly with PBS and fixed for 20 min with freshly prepared 4% paraformaldehyde dissolved in PBS. After permeabilization for 45 min with PBG cells were incubated in blocking buffer (1.5% normal goat serum in PBS). Primary and secondary antibodies were used as described above, followed by incubation with Texas Red Avidin D (1:500; Vector Lab) for 30 min at room temperature. Cells were stained with DAPI and mounted with DABCo. Cells were regarded as positive if at least one focus was visible using a 40× objective in a ZEISS LSM 510 Meta confocal microscope. For quantitative evaluation, 100–300 nuclei were scored per condition.

The antibodies used and the dilution factors were as follows: anti-γ-H2A.X IgG (S139) (#9718, rabbit monoclonal, 1:250; Cell Signaling, Herts, UK), anti-Ki-67 (#M7249, rat monoclonal, 1:50; DAKO, Ely, UK) and anti-8-oxoG (#N45.1, mouse monoclonal, Japan Institute for the Control of Aging, Japan, 1:100).

Apoptotic cells were detected using Apoptag kit (Upstate, Herts, UK) according to the manufacturer’s instructions.

For Sen-β-gal staining, 7 μm mouse liver sections were air dried for 15 min then fixed in 0.4% paraformaldehyde made up in PBS pH 7.2–7.4 for 1 min at room temperature. The sections were washed 2 × 3 min in PBS then stained at 37 °C for 24 h in Sen-β-Gal staining solution containing 2 mm magnesium chloride, 150 mm sodium chloride, 40 mm citric acid, 12 mm sodium phosphate dibasic, 5 mm potassium ferrocyanide, 5 mm potassium ferricyanide and 1 mg mL−1 5-bromo-4-chloro-3-inolyl-β-d-galactoside (X-Gal) at pH 5.5. The sections were washed 2 × 3 min with PBS, counterstained with Nuclear Fast Red solution, washed, dehydrated, cleared in xylene and mounted with DPX.

Telomere Q-FISH and immuno-FISH

Q-fluorescence in situ hybridization was performed as described (Lechel et al., 2007). For immuno-FISH, γ-H2A.X was detected by IF. Following antibody incubations as described above, sections were incubated with avidin–DCS (diluted to 1:250; Vector Lab) for 30 min. Subsequently, tissue sections were washed with 0.1% Tween in PBS and fixed in acetic acid:methanol (1:3) for 15 min and dehydrated with 70%, 90%, 100% ethanol for 3 min each. Slides were air dried, incubated for 5 min at 37 °C in PBS and then fixed in 4% paraformaldedyde for 2 min at 37 °C. Sections were again dehydrated, air dried and then denatured for 5 min at 80 °C in hybridization buffer [70% formamide (Sigma), 25 mm MgCl2, 1 m Tris pH 7.2, 5% blocking reagent (Roche, Welwyn, UK)] containing 25 μg mL−1 Cy-3 labelled telomere specific (CCCTAA) peptide nuclei acid probe (Applied Biosystems, Foster City, CA, USA), followed by hybridization for 2 h at room temperature in the dark. The slides were washed three times for 10 min with 70% formamide, following by three times 5 min wash with 0.1% Tween in PBS. Nuclei were stained by DAPI for 10 min and mounted with DABCo. Analysis was performed in confocal mode in a ZEISS LSM 510 Meta. Pinhole size was maintained at a constant value of 1.2 Airy units to give an optical depth of 1 μm throughout. One hundred to 200 nuclei were scored for each tissue.

For co-localization analysis, confocal images with background thresholds set constant over the whole experiment were deconvolved and analysed using Costes approximation method in imagej, v1.37a (http://rsb.info.nih.gov/ij/). For each nucleus, identified by DAPI staining, a Pearson’s correlation coefficient (R) was determined between the fluorescein isothiocyanate and Cy3 images.


We thank Prof. K.L. Rudolph, Ulm, for providing us with tissue samples from TERC−/− mice and Dr Joao Passos for expert experimental help. This study was supported by programme grant 252 from Research into Ageing and by a BBSRC systems biology grant (CISBAN).