Accumulation of mutations with age has been studied extensively in mice and humans. The early studies of mutations in the HPRT locus in cultured lymphocytes from young and old individuals have shown accumulation of mutations with age in both humans and mice (2–6). The studies using transgenic mouse models allowed the measurement of the mutation frequency in other mammalian tissues and loci. These assays measure mutation frequency in chromosomally integrated LacZ (7) or LacI (8–10) transgenes, which are rescued in Escherichia coli and analyzed for mutations using beta-galactosidase assay. Using these mice, it was demonstrated that point mutations accumulate with age (9–13) and furthermore, the mutation rate is higher in old animals (10). Not only did mutations accumulate but a characteristic type of mutations, genomic rearrangements, appear in old individuals (11,14–18).

Age-related changes in BER have been intensively studied using several experimental approaches. Measuring the levels and kinetics of AP sites following DNA damage in nuclear DNA (32) showed that senescent human fibroblasts as well as leukocytes from old donors have higher basal level of AP sites than young cells. Upon treatment with H₂O₂ or MMS the level of AP sites rose faster in the young cells than in the old cells suggesting a deficiency in DNA glycosylase activity (32).

Measurement of oxidized guanine (8-oxoG) kinetics in genomic DNA of young and old mice after exposure to γ-irradiation did not reveal a difference in the rate of 8-oxoG removal; however, the tissues of old animals appeared to accumulate higher levels of 8-oxoG after irradiation (23). Another way of measuring BER is by comparing in vitro incision activity in tissue extracts from young and old animals. Protein extracts are incubated with radiolabeled oligonucleotides containing single base lesions and the appearance of cleavage products is analyzed on a gel (33). A significant decrease in the mitochondrial incision activity of 8-oxoG DNA glycosylase, uracil DNA glycosylase and the endonuclease III homolog was found in the brains of old mice, whereas smaller changes were observed in nuclear incision activity (30,33,34). Similarly, a decline in cleavage activity was observed in mitochondrial, and to a lesser extent nuclear extracts from senescent human fibroblasts (35).

A different in vitro assay measuring the repair of a synthetic DNA substrate containing a single G:U mismatch showed a strong decline in BER activity in brain, liver and germ cell nuclear extracts of old mice (36). Germ cell extracts from old mice were found to contain reduced levels of APE1, and supplementation with purified recombinant enzyme restored the activity to the level of young animals (37). Reduced abundance of DNA polymerase β in brain extracts from mice and rats has been reported (36,38,39).

The relationship between NER and aging has long been studied by either evaluating repair during in vitro senescence of human fibroblasts and lymphocytes, or in cells or tissues derived from donors of different age. Studies using in vitro senescence yielded conflicting results. Hart and Setlow (45) reported a reduction in unscheduled DNA synthesis following DNA damage in late passage WI-38 human fibroblasts compared to early passage cells. While no differences in the repair replication were found using the same cell line (46). A more recent study (47) measured the kinetics of the disappearance of cyclobutane pyrimidine dimers (CPDs) from genomic DNA in human fibroblasts and trabecular osteoblasts aged in vitro. In this method (48) cells are treated with UV, genomic DNA is extracted and incubated with T4 endonuclease V, which cleaves the DNA at pyrimidine dimers. Then the DNA is cleaved with restriction enzymes, separated on an alkaline gel and the intensity of the bands corresponding to specific genes is determined by Southern hybridization. Both actively transcribed and inactive genes were analyzed, and no clear differences in the rate of CPD removal were found (47). Removal of CPDs was also examined using in situ assay with CPD antibodies (49). This study showed reduced UV-induced CPD removal in senescent compared to young fibroblasts (49).

Contrasting results between different studies may be explained by the sensitivity of cells to culture conditions such as serum concentration, and differences in cell cycle distribution during treatment (50). Furthermore, replicative aging may vary between cell lines, depending on the cell donor. For example, a decline in NER with increasing passages was documented for lymphocytes derived from two different donors but not in the lymphocytes from the third donor (51).

Studies of NER in cells or tissues from young and old individuals consistently showed a decline of NER capacity with increasing age. Earlier works used unscheduled DNA synthesis (52) or a plasmid reactivation assay as a measure of NER (53–55). In the plasmid reactivation assay a plasmid containing chloramphenicol acetyltransferase (CAT) is irradiated with UV to introduce DNA damage, plasmids are transfected into host cells and percentage of CAT activity relative to undamaged control plasmid is measured. A large study using human peripheral blood lymphocytes from 135 individuals aged 20 to 60 showed a decline of UV damage repair (54,55). The rate of decline was calculated as 0.63% per year, which amounts to ~25% decrease over a 40-year period (54,55). This analysis was further extended to show similar decline in NER in skin fibroblasts (53). In addition to reduced plasmid reactivation, cells from older donors introduced an increased number of mutations in the transfected plasmid (53). This suggests that not only the repair becomes less efficient with age, but it also makes more errors. A disadvantage of plasmid reactivation assay is that it does not differentiate between different types of repair, and types of photoproducts.

Other assays measured the rate of removal of specific photoproducts. A study using T4 endonuclease V for detection of CPDs in hepatocytes of 6 and 24 months old rats showed that removal of CPDs from two nontranscribed genes was ~40% lower for cells isolated from old rats than for cells isolated from young animals (56). In contrast, the age-related decline of CPD removal was less apparent in a transcribed gene, where only the rate of CPD removal was slower in old animals, but no difference in the number of CPDs was found after 24 h (56). By a similar assay, repair efficiency of telomeric DNA was reported to be lower in fibroblasts isolated from older human donors (57). Thus, it appears that aging has a greater effect on repair of nontranscribed genes.

Another method of detection of CPDs and (6–4) pyrimidone photoproducts (PP) became available with development of CPD and (6–4) PP-specific antibodies (58). DNA is isolated from cells after UV-irradiation and hybridized with antibodies. In agreement with earlier results this method showed a decrease in removal of CPDs and (6–4) PPs in dermal fibroblasts of older human donors (59). Decline in repair of (6-4) PPs was also shown in round spermatids of 14 months old compared to 2 months old mice (60). The method was further developed to use PP-specific antibodies for immunohistochemistry, which allowed studying the removal of CPDs in situ in human epidermis (61). Skin at the upper arm of young and old volunteers was exposed to UVB, and biopsied at different time points after irradiation. The biopsy material was either used for DNA extraction and probed with CPD antibodies or analyzed by immunohistochemistry. CPDs were removed from epidermis at 4 days after irradiation in the young subjects, and between 7 and 14 days in older subjects (61). Since the process of CPD removal is a combination of NER and epidermal turnover, the CPDs are likely to be removed from aged skin after 7–14 days by turnover of the epidermis. Thus, in situ studies further suggest that NER declines with age. Another in situ study with human volunteers using a postlabeling method based on quantification of photoproducts by HPLC found that photoproducts in UV-irradiated skin are induced at a higher frequency in old individuals (62). The latter study, however, did not detect age-related differences in NER due to large individual variations in the study population (63).

In summary, there is compelling evidence that NER declines with age. Little is known about the mechanism responsible for this decline. Reduced constitutive protein levels of ERCC3, PCNA, RPA, XPA and p53 that participate in NER were reported in older humans (59). Interestingly, treatment with short oligonucleotides that mimic DNA damage signal, stimulated NER in fibroblasts from old donors (64). The oligonucleotide treatment also upregulated the levels of NER proteins (64). Thus, the decline may be caused by lower levels of NER enzymes or by altered induction of DNA damage response.

DSBs in DNA are repaired by two major mechanisms: homologous recombination (HR) and nonhomologous end joining (NHEJ). During HR-mediated repair of DSB, the sister chromatid is used as a template to copy the missing information into the broken locus. Repair by HR is mediated by Rad51 protein with the help of other members of Rad52 epistasis group (68,69). Since sister chromatids are identical to each other, DNA damage can be repaired faithfully with no genetic consequence. In contrast, the NHEJ pathway simply fuses two broken ends with little or no regard for sequence homology. NHEJ starts with binding of Ku70/Ku80 heterodimer to the broken DNA ends (70). Ku facilitates recruitment of Artemis-DNA–PKcs complex, which processes the ends to prepare them for ligation (71). Next, the gaps are filled by DNA polymerase of polX family, and covalently joined by XRCC4-DNA ligase IV complex (71). NHEJ is rarely error-free leading to deletions or insertions of filler DNA (72–74). DSBs situated between two direct repeats can also be repaired by single-strand annealing (SSA), a highly mutagenic recombinational mechanism in which the sequence between the repeats is deleted.

The first indication of age-related changes in DSB repair is the exponential increase in the incidence of cancer with age (75), since cancer is associated with genome rearrangements and loss of heterozygosity (LOH). Sequence analysis of genomic rearrangements in the LacZ locus in aged transgenic mice (11) and also in the endogenous HPRT locus in human lymphocytes (76) did not reveal regions of extended homology at the breakpoints, suggesting that the rearrangements may have resulted from erroneous NHEJ.

Despite the tremendous impact of DSB repair on aging process, age-related changes in DSB repair are currently understudied. Early studies using variations of the comet assay documented age-related decline of the efficiency of rejoining of X-ray-induced DNA breaks in normal human lymphocytes (77,78). The methodology, however, did not allow addressing a specific repair mechanism.

Our laboratory has designed a sensitive assay for NHEJ (73). The assay is based on a reporter cassette containing GFP gene interrupted by an intron, containing recognition sites for the rare cutting endonuclease I-SceI. The GFP gene in the original construct is inactive, however, induction of DSBs by I-SceI and subsequent repair restores GFP activity. Since NHEJ takes place within intronic sequences, GFP activity is reconstituted even if repair is accompanied by deletions or insertions. The number of GFP positive cells serves as a measure of NHEJ efficiency, and can be quantified by FACS. In addition the fidelity of repair can be determined by sequencing the repair products. We used this assay to study senescence-related changes of NHEJ in normal human fibroblasts. The efficiency of NHEJ was reduced up to 4.5-fold in presenescent and senescent relative to young cells. Strikingly, we observed that end joining in old cells was associated with extended deletions. These results indicate that end joining becomes less efficient and more error-prone in senescent cells (73).

NHEJ in the brain of young and old rats has been analyzed using in vitro plasmid rejoining assay. In this assay linearized plasmids are incubated with nuclear extract, and plasmid concatamers resulting from NHEJ are quantified on a gel. This assay showed that NHEJ efficiency is reduced in the brains of old rats (79,80). NHEJ is also decreased in the brains of Alzheimer's disease patients compared to normal controls (81).

What is the molecular mechanism responsible for age-related decline of NHEJ? The level of Ku, which is a protein responsible for recognition and binding to DSBs, has been shown to decline markedly in the testis of aging rats (82), and lymphocytes from human donors of increasing age (83). We observed a 50% reduction in Ku70 and Ku80 protein levels in senescent human fibroblasts (84). Furthermore, intracellular distribution of Ku differed between young and senescent cells. In young cells, Ku was distributed throughout the cytoplasm and the nucleus, and translocated to the nucleus in response to γ-irradiation. We hypothesize that the cytoplasm of young cells contains Ku that serves as a reserve that is recruited to the nucleus upon DNA damage. In contrast, in senescent cells, all Ku was localized in the nucleus, so no additional Ku can be brought into the nucleus to repair DNA damage. Moreover, our results suggest that nuclear Ku in senescent cells is unavailable for repair of new lesions (84). Ku may be trapped by unrepairable DNA damage foci (21,22) that accumulate in the nuclei of senescent cells. Ku heterodimer forms a ring that is loaded onto the DNA strand (85), and it is unknown what takes Ku off the DNA. The release of Ku from DNA may be impaired in senescent cells. Alternatively, Ku may remain at the damage sites and shortened telomeres, which cannot be repaired by DNA repair machinery in senescent cells. Impaired nuclear targeting and DNA binding activities of Ku have been reported in peripheral blood mononuclear cells of aged humans (86,87). Thus, decline of NHEJ with age may be caused by reduced availability and altered regulation of Ku.

Can this finding be generalized to mammalian DNA repair? We did observe 4.5-fold decrease in NHEJ in senescent human fibroblasts (73), however, this was clearly not due to an increase in HR, since HR is virtually absent in replicatively senescent cells (Seluanov,A. and Gorbunova,V., in preparation). Lack of HR in G0-arrested cells is not unexpected, as HR is restricted to G2/M phase of cell cycle when a sister chromatid is available (90,91). Potentially, a homologous chromosome can be used as a template in nondividing cells, however, repair using a homolog is three orders of magnitude less frequent than the use of a sister chromatid in mammalian cells (92,93). This preference is explained by sister chromatids being in close proximity after DNA replication, in contrast to homologs that are not paired in mitosis and may be harder to bring together. Expression of Rad51 protein is cell cycle-regulated (94,95), and Rad51 becomes undetectable when cells enter replicative senescence (Seluanov,A. and Gorbunova,V., unpublished data). Therefore, the major pathway for repair of DSBs in mammalian G1/G0 cells is NHEJ (90,91,96).

While senescent cells do not use HR, the situation is less clear in an aging organism, which still contains dividing cells. However, several studies show that HR is more active in ES cells (97) and in early embryonic development (98), compared to late developmental stages and somatic cells. As the body ages, fewer cells divide, cell cycle slows down and more cells are in G1/G0 stage, therefore we predict that the usage of HR will further decrease. This prediction, of course, needs to be tested in aging mammals.

Do the observed changes in DNA repair activities with age contribute to the onset of aging? It may be difficult to infer cause–effect relationship from the studies of normal aging, however, mouse knockout studies can provide answers. Defects in DNA maintenance may lead to accelerated aging. Notable examples are mice deficient in NHEJ or mice deficient in TCR (107,108). It has to be mentioned that not all DNA repair mutants show progeroid phenotypes, and it appears that only mutations in particular pathways affect lifespan. However, as most premature aging syndromes are caused by mutations in DNA repair genes it is reasonable to assume therefore that normal aging is caused, in part by the decline in DNA repair capacity.

The overall sequence of events can be as follows (Figure 2): spontaneous mutations and rearrangements gradually impair the function of genes involved in stress response and DNA repair. DNA repair becomes less efficient and more error-prone leading to cascading accumulation of DNA damage and mutations, which further exacerbate age-related physiological decline.

Accumulation of mutations may contribute to aging by affecting essential genes and disturbing transcription patterns (1). In addition, unrepaired DNA damage may contribute to aging by increasing apoptosis. A shift in the balance between cell death and cell renewal may lead to exhaustion of the stem cell pool, loss of tissue cellularity and functional decline. Accelerated aging due to loss of stem cells has been observed in mice with overactive p53 (109). Mice with mutated DNA polymerase γ lacking the proofreading activity exhibit accelerated aging due to increased apoptosis (109), and remarkably several NER mutants undergo premature aging due to increased apoptosis, which is not associated with an increase in mutation frequency (110). Furthermore, DNA damage and mutation may perturb chromatin structure and cause epigenetic changes.