Structural chromosome abnormalities, increased DNA strand breaks and DNA strand break repair deficiency in dermal fibroblasts from old female human donors

Age-related regulation of genes associated with genome maintenance, cell cycle and DNA-repair

Aging is an inevitable and ubiquitous development in living organisms, while age progression is an individual feature and shows a large heterogeneity among humans. To that effect the fibroblast transcriptomes from the differently aged donor groups studied here hinted at a heterogeneous response as indicated by principal component analysis (Fig. 1A). Pronounced heterogenei-ty of the middle-aged group was previously assigned to the increased body mass index of this group [20]. Here, we show that cells from old donors were also heterogeneous. Some clustered apart from the young donors (samples from donors aged 60, 63 and 67), while others clustered together with the young donors (samples from donors aged 62 and 64). Correspondingly, gene set enrichment analysis (GSEA) on major age-associated pathways revealed a large variation between young and old donor samples. Generally, we observed a down-regulation of gene sets encoding for cell cycle, extracellular matrix and translation [20]. These findings were in part corroborated by corresponding changes in the proteome [6]. To differentiate these diverse functional alterations and identify age-associated transcriptome alterations that are specifically related to genome stability, we here performed a more detailed analysis on gene sets associated with telomere maintenance and -repair, chromosome segregation and -maintenance as well as DNA double-strand repair (See Supplementary Table I). For the old donors that were distinct from the young donors (samples 60, 62 and 64) this analysis suggested an age-related decrease of transcription/translation and cell cycle that was accompanied by impaired telomere and chromosome maintenance (Fig. 1B). However, there was no clear significance for a concordant regulation of DNA repair genes with age. Any expression changes of DNA repair genes in the samples of old donors as compared to young donors appeared to be covariant with up- or down-regulation of cell cycle genes.

Age-correlated loss of proliferative capacity and increase in heterochromatin marks

It has been pointed out that loss of proliferative capacity is a primordial feature of cell ageing [33]. Indeed, we observed that cell proliferation steadily and significantly decreased with donor age (Fig. 2A), which conforms to age-associated down regulation of genes related to cell cycle progression and mitosis demonstrated in Fig. 1B and Suppl. Tab. 1. The decrease in cell proliferation is not likely related to replicative telomere shortening and replicative senescence, because we have previously shown that replicative cell cycle arrest does not occur below 40 population doublings (PD), and that induction of replicative senescence in culture can be avoided by keeping the cells well below that limit [20]. Moreover, telomere length determined under these conditions exhibited an insignificant decrease of only about 0.01 kb per year of donor age and was on average around 3.6 kilo base pairs (Waldera-Lupa et al., currently under revision in J Invest Dermatol), which is clearly above the limit considered critical for telomere function [34].

Thus, dermal fibroblasts appear to undergo replicative telomere shortening only to a very minor extent during ageing in situ, which conforms to current concepts on skin ageing [1, 2]. Consequently, age-related changes of cell functions observed here under these conditions are not likely related to the induction of replicative senescence in vivo or in vitro. In skin of aged primates heterochromatic marks such as heterochromatin protein 1β (HP1β) and histone macroH2A (mH2A) are increased [10], and human fibroblasts subjected to replicative senescence in vitro undergo progressive conversion of eu- into heterochromatin [28]. Here, we observed a significant increase in the expression of mH2A in still replicating fibroblasts from old donors, and a similar trend in the expression of HP1β (Fig. 2B), which indicates that age-related increases in heterochromatin marks can occur independently of replicative senescence. In summary these observations, confirm the concept that cell ageing encompasses two independent processes: loss of proliferative capacity and geroconversion, that is, the acquisition of irreversible age-associated functional alterations such as increased heterochromatinisation [33].

Increases in chromosome breakage and DNA damage response at base line

Structural and numerical chromosome aberrations are known to increase with age in peripheral nucleated blood cells, buccal epithelia [35–39], hepatocytes [40], vascular smooth muscle cells [41] and human brain [42, 43]. Human fibroblasts subjected in culture to replicative or stress-induced senescence accumulate γH2AX foci [44], chromosome- and centrosome aberrations [24, 26], and exhibit a decline in pathways for rapid repair of DNA double strand breaks (DSB) [25]. A decline of DSB repair capacity was also observed in nucleated peripheral blood cells of old human donors [45]. Based on these reports and in the light of the down regulation of genes associated with genome maintenance in the cell samples studied here (Fig. 1B), we hypothesised that geroconversion occurring in dermal fibroblast during ageing in situ could encompass the onset of chromosome instability.

To address this question, we subjected the cells to classical cytogenetic analysis. We observed a high incidence of non-clonal chromosome aberrations. About 20% of the mitoses were abnormal and exhibited a variety of abnormalities (Fig. 3B). On average, we found more than one chromosomal aberration per abnormal mitosis (Fig. 3C). The overall frequency of chromosome abnormalities was not sign ificantly correlated to donor age. However, upon differentiation between structural and numerical chromosome aberrations (examples see Fig 3A), we were able to detect a significant age-correlated increase in the incidence of structural aberrations suggesting enhanced chromosome breakage and translocation (Fig. 3D). In contrast, numerical aberrations indicative of mitotic dysfunctions did not increase with donor age (Fig. 3E). These observations suggested that ageing of dermal fibroblasts analysed ex vivo involves an increase in chromosome and chromatid breaks.

To address whether the age-related increase in chromosome breaks is associated with an increase in base line levels of DNA doub le-strand breaks (DSB) we quantified histone 2AX phosphorylated at serine 239 (γH2AX), which is a robust quantitative parameter for the DSB-associated DNA damage response (DDR) [46].

To compare base line level and maximal capacity of DSB-associated DDR, γH2AX was assessed with and without exposure to the radiomimetic drug etoposide (VP16), which was applied at a dose (50 μM) known to saturate DSB-elicited DDR in human cells [46]. Base line levels of γH2AX showed an age-related exponential increase, while the maximal response levels of γH2AX to 50 μM VP16 did not increase accordingly. As a consequence, the amplitude of the γH2AX response to VP16 declined from about 100-fold in cells from donors aged 20–30 years to less than 10-fold in cells from donors aged 60–70 years (Fig. 4A, please note log scale of y-axis). Moreover, across all cells tested, average base line levels of γH2AX showed a reasonable correlation with the percentage of cells bearing structural chromosome aberrations (Fig. 4B). These data suggest that age-related increases in chromosome breaks could indeed be due to an increase in base line DSB.

Increased levels of residual DNA breaks at base line

To corroborate that notion and distinguish whether the age-associated increase in γH2AX (Fig. 4A) was a primary alteration of DDR or a response to increased DSB, we directly assessed overall levels of DNA breaks. This was done by fluorimetric detection of alkaline DNA unwinding (so called FADU-assay) [45, 47, 48], which was measured before and immediately after exposure to ionising radiation (IR, 3.8 Gy). As shown in Fig. 5A, the extent of residual alkaline DNA unwinding before IR exposure increased with age, whereas the extent of additional DNA unwinding induced by IR was the same in cells from old and young donors. The latter observation indicates that the IR-susceptibility of the nuclear genome did not change during ageing, which was confirmed by the stringent overlap of IR dose-response curves of all donors (Fig. 5B). Therefore, the age-related increase in alkaline DNA-unwinding at base line seems due to an increase in residual DSB load unrelated to IR-sensitivity, and that increase could be the reason for the increase in base line γH2AX (Fig. 4A) and the spontaneous increase in chromosome breakage (Fig. 3D). The latter notion is supported by reasonably good correlations between residual alkaline DNA unwinding and base line γH2AX (Fig. 5C) as well as between residual alkaline DNA unwinding and the load of structural chromosome aberrations (Fig. 5D) across cells from all donors. In summary these data strongly suggest that age-correlated increases in base line DNA damage signalling and structural chromosome abnormalities are due to an increase in residual DSB load. However, it remained unclear why residual DSB load increases with donor age.

Heterogeneous alterations of DSB-repair in old cells

One possible explanation for the observed increase in residual DSB load would be a decrease of repair capacity. However, age-related down regulation of genes involved in telomere maintenance and -repair, chromosome segregation and chromosome maintenance (Fig. 1B), was frequently associated with up regulation of genes involved in DSB repair. This was most conspicuously the case in the samples from donors aged 62/64, which showed reduced expression of genes involved in telomerase regulation and the telomere-associated SLX4 complex, but at the same time exhibited increased expression of genes playing a prominent role in non-homologous end-joining (NHEJ). This combination is suggestive of a decreased efficiency in chromosome maintenance, which is partially balanced by increased DSB repair by NHEJ. We used the FADU assay to determine the kinetics of disappearance of IR-induced DNA strand breaks, which is considered a valid measure of DSB-repair activity and capacity [45]. Cells were challenged with 3.8 Gy, a dose inducing a half-maximal increase in alkaline DNA unwinding (Fig. 5B). Disappearance of that increase was then monitored over 90 minutes. The data were normalised to the initial IR-induced increase in alkaline DNA unwinding and plotted over time. Fig. 6 demonstrates these repair trajectories for the three age groups. It becomes apparent that the cells from young donors exhibited very similar, mostly overlapping repair trajectories resulting in the disappearance of 40 - 80 % of the IR-induced DNA unwinding (Fig. 6, left, open circles) within 90 min. Repair trajectories of the middle aged donor group (Fig. 6, middle, semi-closed circles) were more heterogeneous with endpoints ranging between 10–90 % disappearance of IR-induced DNA unwinding. Repair trajectories of the old donor group (Fig. 6, right, closed circles) appeared even more heterogeneous: In three specimen repair trajectories were similar as in the middle aged group with endpoints ranging between 40–95 % disappearance of IR-induced DNA unwinding, while the other two specimen exhibited the opposite, namely a progressive increase in DNA unwinding that went beyond the extent initially induced by IR (i.e. an aggravation of DNA damage over time). These results are in line with the gene expression heterogeneity depicted by the PCA and GSEA in Fig. 1: Two samples from old donors (63 and 67) clustered together with the middle aged and young donors. These cells exhibited a higher expression of genes related to cell cycle and DSB repair and had a similar or even better capacity to repair IR-induced DNA damage than the young group. In contrast, the other three specimen of the old group (60,62 and 64), which clustered separately from the young and middle-aged donors, exhibited a decreased expression of genes related to cell cycle and DSB repair, and were less efficient or even unable to repair IR-induced DNA damage (compare Fig. 1A and Fig. 6).

Thus, in some cases repair of DNA strand breaks became more efficient with age, whereas in other cases it became less efficient or even completely dysfunctional, and it remains unclear what determined this divergent outcome. It seems unlikely that age-correlated regulation of DNA repair genes is primarily involved, since, in line with the repair trajectory experiments (Fig. 6), GSEA revealed a heterogeneous age-related regulation of DNA repair related gene sets (Fig. 1B). Only 32 out of the 405 genes annotated with DNA repair processes were differentially regulated with age (Fig. 7A, robust linear regression. P-value < 0.05, R² > 0.4). Please note that four donors from the middle aged group having a large body mass index and being outliers in the PCA (Fig.1A) were excluded from that analysis. Among the few repair genes consistently regulated with age, the down-regulated genes (CEP164 and TP53) are also related to the cell cycle. Thus, down regulation is probably associated with the age-related drop in proliferation demonstrated in Fig. 2B. In contrast, age-regulated genes only involved in DSB response and repair (EYA4, XRCC4 and LIG4) were up-regulated with age. Most notably, XRCC4 and LIG4 showed a linear, age-correlated increase in expression (Fig. 7B). These two genes code for rate limiting components of NHEJ, the major DSB-repair system available to human dermal fibroblasts at their normal G₀ state [49]. However, despite the uniform, age-correlated up-regulation of these DSB-repair genes, some of the old cells failed in the repair of IR-induced DNA damage, while others excelled.

One possible explanation for the above divergence could be differences in heterochromatin formation. We have demonstrated in Fig. 2A that expression of heterochromatic marks increased with donor age. It is long known that DNA-damage in heterochromatin is refractory to repair [50]. Moreover, primarily un-repairable DSB induce adjacent heterochromatin formation [51]. However, when comparing the expression levels of heterochromatin marks shown in Fig. 2B with the repair trajectories shown in Fig. 6, it turned out that the cells with the highest levels of HP1β and mH2A expression (i.e. donors 20 and 67) exhibited the highest capacity to repair IR-induced DNA damage, which argues against a dampening effect of age-associated heterochromatinisation on the efficiency of these cells to repair IR-induced DNA breaks.

Concluding remarks

Our data suggest that during ageing in situ dermal fibroblast progressively lose their ability to control base line DSB levels, which leads to increased DNA recombination, chromosome breakage and enhanced background noise of DDR signalling. Human fibroblasts subjected to replicative or stress-induced senescence in culture are known to accumulate DNA damage foci believed to represent unrepaired DSB [44] or unrepaired DNA damage at telomeres [27]. Similarly, lymphocytes from aged humans exhibit increases in base line DNA breakage [45, 52, 53] and chromosomal aberrations typically triggered by imprecise repair of DSB via the NHEJ pathway [35, 37]. This type of age-related genome instability is clearly different from the increase in polyploidy observed in aged vascular endothelial cells [54], which is related to telomere shortening and replicative senescence but not to increases in DSB [55]. The question is, whether the age-related increase in DSB observed in fibroblasts and lymphocytes is due to an increased rate of damage or a decreased rate of repair.

There are many indications that age-related increases in residual DSB load and chromosome instability are linked to alterations in DSB repair capacity. NHEJ is probably the only DSB-repair system available to dermal fibroblasts, because in the skin these cells are mostly in G₀ phase and therefore lack a template for repair by homologous recombination [49]. NHEJ becomes less efficient and less precise, when human fibroblasts enter replicative senescence [25], although replicative senescence of dermal fibroblasts induced in vitro is not associated with a significant slowing of removal of IR-induced DNA damage [56]. NHEJ activity is also reduced in brains of old rats and Alzheimer disease patients, and it declines in peripheral human lymphocytes with donor age [45, 52, 53, 57]. However, it remains unclear whether age-associated genome instability is the cause or the consequence of age-associated alterations of DSB repair.

Here, we observed that dermal fibroblasts from old human donors have a heterogeneous capacity to remove IR-induced DNA strand breaks. Since IR mostly induces DSB, the observed inter-individual differences in removal of IR-induced DNA damage are most likely due to changes in the efficiency of DSB-repair. Given that gene expression of pace-making elements of the NHEJ pathway showed a uniform age-related increase, the heterogeneous functional outcome of DNA repair possibly indicates the overlay of adaptive and mal-adaptive processes in these long-lived cells. Abrogation of DSB-repair could also be due to accumulation of un-repairable lesions, which has been observed in various other ageing models [27, 44, 58]. A third possibility would be that age-related increases in heterochromatin formation impede DSB-repair in some of the old cells but not in others. However, our data argue against this possibility because all cells showed a similar age-correlated increase in heterochromatin marks. In summary, these considerations prompt the conclusion that the observed alterations of DNA repair capacity reflect a diverse response of age-associated increases in DNA damage and not a cause thereof. In line with that conclusion, a recent systematic analysis of the microRNA-transcriptome of the same cell samples as studied here has revealed a panel of at least 12 microRNAs that exhibit significant age-related up- or down-regulation, but none of the 164 mRNA targets predicted for these microRNAs was related to DNA-repair [6].

Common belief holds that cell cycle arrest is a common feature of senescent and quiescent cells, while cellular senescence emerges from cellular quiescence by the irreversible process of geroconversion, which is driven by p53- and mTOR-signalling and entails the loss of proliferative potential and the acquisition of cellular hallmarks of aging [33]. A similar increase in DNA damage signalling as seen here in dermal fibroblast from old humans was previously observed in dermal fibroblasts of old baboons and ascribed to the accumulation of senescent cells in the skin of the animals [8]. In our ex vivo study we monitored the cells in proliferating culture. Since senescent cells stop to proliferate, they should not accumulate in such a culture and therefore should not contribute significantly to the age-related phenotype changes observed therein. Our data therefore support the alternative concept that in long-lived dermal fibroblasts geroconversion occurring in vivo entails the onset of genome-instability as a primary feature, which is stably maintained when the cells are brought back into the cell cycle and therefore unrelated to cellular senescence.

Due to the small cohort size our study is at best a piloting investigation, which, moreover, has been restricted to cells from female donors. Gender and estrogen deficiency at menopause are known to have a much stronger impact on gene expression in skin that chronological age [32]. On the other hand, gender was found to have no influence on DNA damage and DNA repair capacity in human lymphocytes [45]. Thus, it remains an open question whether our results also apply to aging processes in male skin. Clearly, studies on a broader base of biological specimen from both genders are required, to obtain a comprehensive picture of chromosome abnormalities, DNA strand breaks and repair deficiency in dermal human fibroblasts in relation to age.