Delayed γH2AX foci disappearance in mammary epithelial cells from aged women reveals an age-associated DNA repair defect

Defining the criteria for analyzing DNA double strand breaks in pre-stasis HMECs

HMECs were obtained from reduction mammoplasty tissue of 12 donors, which were classified according to age into young donors (YDs, ≤ 27, age in parentheses): YD48R(16), YD240L(19), YD168R(19), YD184(21), YD59L(23) and YD123(27) and aged donors (ADs, ≥ 60, age in parentheses): AD153L(60), AD112R(61), AD122L(66), AD29(68), AD429ER(72), AD353P(72). Cells were cultured as pre-stasis strains in M87A medium as described by Garbe and colleagues [21], to support their long-term growth (Figure 1A). Despite using a low-stress medium, there was an accumulation of senescent cells with time in culture (Figure 1A and 1B). In order to avoid interference from replicative-senescence associated DNA damage when assessing age-dependent differences in the formation and resolution of DSBs, early PDs were chosen (PD < 20 which correspond to passages 4^th to 6^th) in which the frequency of senescent cells was ≤ 10%.

As previously reported [21,22] we found age-related differences in the fractions of myoepithelial (CD10⁺/CD227^-) and luminal (CD10^-/CD227⁺) cells in HMEC culture. Flow cytometry analysis of CD10 and CD227 cell-lineage specific markers confirmed an age-dependent decrease in the myoepithelial fraction accompanied by an increase of the luminal fraction (CD10⁺/CD227^- in YD240L: 56.05%; AD112R: 37.29%; CD10^-/CD227⁺ in YD240L: 12.67%; AD112R: 20.06%) (Figure 1C). In order to rule out radiation-sensitivity differences between the two breast cell types, cells from young and aged donors were exposed to 1Gy of γ-rays and labelled with γH2AX and claudin-4 (Cl4), a cytoplasmic membrane protein mostly expressed by luminal cells (Figure 1D). As shown in Figure 1E, there were no differences in the frequency of γH2AX foci between Cl4⁺ and Cl4^- cells 2h after irradiation in any of the donors analyzed (Cl4^-: 31.89 and Cl4⁺: 33.42 in the YD184; Cl4^-: 27.68 and Cl4⁺: 26.41 in the AD112R; Mann-Whitney test, p-value > 0.05). These results indicate that radiation-induction of DSBs is similar in myoepithelial and luminal HMECs, ruling out the need to distinctively identify them when analyzing age-dependent differences in DNA repair.

Mammary epithelial cells from aged donors show an increased basal frequency of DSBs

γH2AX foci are accepted as surrogate markers of DSBs [23], but the pattern of γH2AX staining and the number of foci scored are dependent on the phase of the cell cycle analyzed (Supplementary Figure 1). To mitigate variability due to cell cycle, γH2AX foci counting was restricted to cells in G1 phase, which were identified by pericentrin labelling, a centrosomal protein that duplicates along with DNA, allowing clear distinction of cell cycle phase for each individual cell analyzed [24]. γH2AX foci were scored before and after exposure of HMECs to IR (1h, 2h and 24h pIR). In order to detect differences in γH2AX foci numbers between age groups (young donors vs old donors), a generalized linear model with repeated measures for each donor was established (see Materials and Methods section). We first established the basal frequency of DSBs in cells from young and old donors, in non-irradiated samples. Using the generalized linear model, the estimated mean number of γH2AX foci before irradiation was 0.96 (CI_95% = [0.70, 1.30]) in cells from YDs and 1.94 (CI_95% = [1.43, 2.63]) in cells from ADs, twice the level scored in YDs (Figure 2A). Statistically significant differences between basal γH2AX foci frequencies in YDs and ADs were detected (p-value = 0.0013; t = -3.22). In addition to a lower basal frequency of DSBs, in most young donors (5 out of 6) 60%-75% of cells were devoid of any γH2AX foci, whereas in most old donors (5 out of 6) less than 45% of cells were devoid of foci (Figure 2B). In addition, in most YDs less than 10% of cells carried more than 3 γH2AX foci per cell, whereas ~20% of cells from ADs had more than 3 foci (X² test, p-value < 0.0001). Despite the existence of inter-individual differences between donors of similar ages, these analyses demonstrate that both, the average basal frequency of DSBs and the fraction of cells carrying DSBs are higher in HMECs from aged donors as compared to young donors.

Descriptive statistics was computed for each donor (Table 1 and Supplementary Figure 2) and statistical differences regarding the mean number of γH2AX foci per cell of each donor were calculated (Kruskal-Wallis test with a Dunn’s multiple comparisons correction). When donors were lined up based on statistical differences among them, most of YDs and ADs aligned according to an age-dependent order (Figure 2C Non-irradiated). This analysis allowed us to detect that unirradiated cells from YD123(27) and AD153L(60) did not behave as the rest of the donors of their age group (Table 1, Figure 2C Non-irradiated), thus unmasking the existence of inter-individual differences among donors. Besides these particular exceptions, the rest of YDs had a similar and low frequency of basal DSBs/cell (Table 1 and Figure 2C Non-irradiated) and, consequently, they statistically grouped together (a, b and c) and were significantly different from most of ADs (d, e), which carry more basal DSBs/cell and display a greater data dispersion.

Aged donors accumulate higher levels of DSBs after irradiation

To study the efficiency of DSB repair with age, exponentially growing cell cultures from all donors were exposed to 1Gy of γ-rays. One hour after IR exposure, the estimated mean number of γH2AX/cell was 15.70 (CI_95% = [11.57, 21.29]) in YDs versus 22.27 (CI_95% = [16.42, 30.19]) in cells from ADs (Figure 2A). As shown in Table 1, at this time point the mean number of γH2AX foci per cell strongly correlated with the age of the donors, ranging from 10.63 γH2AX foci per cell in the youngest donor (YD48R(16)) to 32.7 γH2AX foci per cell in the oldest donor (AD353P(72)). Alignment of donors at this time according to statistics (Kruskal-Wallis test and Dunn’s multiple comparisons correction), rendered clear differences between young and old donors and most of them continued to maintain an age-related position (Figure 2C 1h after irradiation). Again, data from YDs showed little variance, revealing similar DSB repair efficiencies while ADs presented with more γH2AX foci and higher inter-cellular variability. Overall, 1h after irradiation cells from ADs accumulated higher levels of unrepaired DSBs, suggesting that these cells elicit a less efficient response from the fast component of DSBs repair.

When γH2AX foci were scored two hours after IR exposure the estimated mean number of γH2AX foci per cell had already decreased in all donors and it was similar for YDs (11.58 γH2AX foci/cell, CI_95% = [8.54, 15.70]) and for ADs (14.08 γH2AX foci/cell, CI_95% = [10.38, 19.08]) (Figures 2A and 2C 2h after irradiation, Table 1). The decline in γH2AX foci during this second hour was higher in cells from ADs than in cells from YDs, suggesting that the initial impairment in DSBs repair shown by ADs 1h after irradiation is eventually alleviated.

In order to evaluate the efficiency in the slow component of DNA repair, we finally analyzed the frequency of γH2AX foci 24 hours after IR exposure. Both YDs and ADs have repaired most of the radiation induced DSBs, but while most of the YDs had reached a frequency of residual DSBs close to the basal levels, only two aged donors had reached their basal levels of DSBs (Table 1). Thus, cells from ADs displayed a higher estimated mean number of γH2AX foci/cell than cells from YDs (YDs: 1.24 γH2AX foci/cell, CI_95% = [0.91, 1.68]; ADs: 2.85 γH2AX foci/cell, CI_95% = [2.10, 3.86]; p-value = 0.0001; t = -3.79) (Table 1, Figure 2A). Indeed, when donors were individually compared (Kruskal-Wallis test and Dunn’s multiple comparisons correction) the differences between YDs and ADs allowed a clear age-related alignment (Figure 2C 24h after irradiation). Not only YD cells present with less γH2AX foci/cell, but also the frequency of cells devoid of γH2AX foci at 24h is 50%, close to their frequency before irradiation (70%) (Figure 2D). In contrast, in ADs the frequency of cells without γH2AX foci at 24h after irradiation is far from their basal frequency (15% vs 40%) (Figure 2E). Among cells with γH2AX foci, most of the YDs’ cells scored only 1 or 2 γH2AX foci per cell at 24h pIR, whereas ADs still accumulated 3 or more γH2AX foci per cell (Figures 2D and 2E). Thus, at 24 hours after irradiation more cells from ADs accumulate DSBs, and also the frequency of DSBs per cell is higher than in YDs.

Finally, and in order to determine if the γH2AX foci disappearance was a good marker of chronological age, we carried out a hierarchical clustering analysis using standardized values of γH2AX foci from the 12 donors in the 4 time points (non-irradiated, 1h, 2h and 24h after IR). With these data, the donors were grouped in 3 clusters (Figure 2F). A clearly separated cluster was constituted by the oldest donors (AD122L(66), AD429ER(72) and AD353P(72)), which displayed the worst repair efficiency among all the donors. The 4 youngest donors (YD48R(16), YD168R(19), YD240L(19) and YD184(21)), which are the ones with the best DSB repair performance, clustered together and separated from the other donors. And finally, an intermediate cluster included the remaining young donors (YD123(27) and YD59L(23)) along with the 3 aged donors (AD153L(60), AD112R(61) and AD29(68)) that frequently did not follow an age-dependent order in the previous statistical analyses (Figure 2C). Hence, hierarchical clustering of donors according to γH2AX foci at different times after irradiation reveals that DSB repair efficiency is a good marker of age.

Delayed firing of the DNA Damage Response (DDR) with age

DSB repair is not constant, as it follows biphasic exponential negative kinetics. In order to determine the nature of the repair defect displayed by cells from older donors, we aimed to describe the kinetics of DSB repair for the two age groups. We first calculated the rate of γH2AX foci disappearance for each time interval analyzed (Table 2). Because γH2AX foci assay does not allow the scoring of the DSBs induced immediately after irradiation (𝜃), to estimate γH2AX foci disappearance at the initial time interval we have used the previously described standard estimation of 35 DSBs induced per Gy of radiation in G1 cells [25]. According to this, during the first hour after DNA damage induction, the rate of DSB resolution was higher for YDs (53.83% of γH2AX foci disappeared) than for ADs (39.51%) (Figure 3A and Table 2), indicating a greater DSB repair ability for YDs immediately after DNA damage induction, while ADs end the 1^st hour carrying higher numbers of unresolved DSBs. In contrast, the rate of γH2AX foci disappearance between 1 and 2h after IR was higher in AD (21.09%) than in YD samples (13.51%). Two hours after irradiation ADs have repaired 60.6% of radiation-induced DSBs, very close to the 67.3% of DSBs repaired by YDs’ cells, suggesting that, although with some delay, cells from ADs are eventually able to launch the DDR and efficiently resolve the accumulated DSBs. The last time interval analyzed (2 to 24h post-irradiation) corresponds to the slow component of repair, in which the rates of DSBs repaired were very similar for both young (28.63%) and aged donors (31.49%) (Figure 3A and Table 2), suggesting that age-related differences in DNA repair efficiency lay within the initial times after DNA damage induction.

In order to test the hypothesis of an age-related delay in DNA-DSB repair initiation, we established a first order kinetic reaction using a Nonlinear Regression Model. The model, described in Materials and Methods section, predicts the number of γH2AX foci in AD and YD cells and it is resolved as follows:

${[n\_foci]}_t=\left\{\begin{array}{c}24.32* e^{-0.44\text{ *}\mathrm{ }t}+ {[n\_foci]}_0 if YD\\ 24.32* e^{-0.44\text{ *}\mathrm{ }t+0.27}+ {[n\_foci]}_0 if AD\end{array}\right)$

The repair of DSBs is estimated by the constant of γH2AX foci decay β₁ = -0.44, CI_95% = [-0.64, -0.25] for young and aged donors. In the case of ADs, the model includes a constant factor of delay in DSB repair initiation, which is estimated to be β₀ = 0.27, CI_95% = [0.08, 0.45]. Thus, the equation is different for YDs and for ADs, as it assumes that ADs present a delayed start of DSB repair (e^-0.44*t+0.27) with respect to YDs (e^-0.44*t). The basal frequency of DSBs ([n_foci]₀) is also included in the equation because it is different for YDs and ADs (Table 1). Finally, the model estimates that the initial number of γH2AX foci in G1 phase HMECs induced by 1Gy of γ-rays is 𝜃 = 24.32, CI_95% = [17.28, 31.35]. Therefore, immediately after irradiation the cells carry those radiation-induced DSBs plus their basal frequency of DSBs ([n_foci]₀) and, according to the model, these γH2AX values are 25.22 for YDs and 26.35 for ADs. Although our model’s estimation of induced DSBs is lower than the 35 DSBs/Gy reported previously [25], it is very similar to others’ estimations of ~25 DSBs/Gy in G1 cells [26,27]. Discrepancies in the number of DSBs induced can be attributed to the source of radiation, the dose rate used in each experiment or to an overestimation of the number of DSBs detected by PGFE methodology.

As shown in Figure 3B, this model renders estimated DSB repair kinetics between 1 and 24h after IR for YDs and ADs that fit well the data observed. Although not strictly applicable, the model has also been used to make an extrapolation corresponding to the kinetics of DSB repair from the time point immediately after irradiation to 1h pIR, shown as dotted colored lines (inset in Figure 3B). The dotted line in YDs reaches the value of 25.22 γH2AX foci at a time close to 0 (blue arrowhead in Figure 3B inset), suggesting that YDs initiate repair immediately after irradiation and they efficiently diminish the number of DSBs during this first hour. Instead, ADs maintain the number of γH2AX foci they had immediately after IR for a longer time, because when data obtained is extrapolated from 1h pIR backward (dotted red line in Figure 3B inset) the value of 26.35 γH2AX foci is attained at a time between 0 and 1h (red arrowhead in Figure 3B inset), suggesting that ADs begin to resolve DSBs later than YDs. Thus, ADs reach the first hour after irradiation carrying more γH2AX foci, which are markers of unresolved DSBs. Hence, it is tempting to speculate that a period of latency exists before ADs are able to fire a fully operative DNA repair response, although once launched, they are able to repair with a speed similar to that of YDs.

Cell culture

Finite lifespan pre-stasis HMECs were obtained from reduction mammoplasty tissue of 11 donors: 48R (16 yo), 240L (19 yo), 168R (19 yo), 184 (21 yo), 59L (23 yo), 123 (27 yo), 153L (60 yo), 112R (61 yo), 122L (66 yo), 29 (68 yo), 429ER (72 yo), or peripheral non-tumor containing mastectomy tissue of 1 donor: 353P (72 yo). Donors were classified into two groups depending on age: young donors (YDs, ≤ 27 years old) and aged donors (ADs, ≥ 60 years old). When referring to donors, the group of age is followed by the specimen identification and the age of the donor in parentheses.

Cells were cultured as pre-stasis strains using M87A medium with cholera toxin and oxytocin according to previously reported methods [21], with the addition of 100 U/ml penicillin and 100 μg/ml streptomycin. Cells were incubated at 37ºC and 5% CO₂ atmosphere. The passage number indicates each time the cells have been detached form the petri dish using trypsin and seeded into new vessels. The population doubling (PD) indicates each time a cell has divided and was calculated as described by Greenwood and colleagues [41], using the equation: PD = [log(cells harvested/cells plated)]/log2. Cells have been obtained from mammary gland surgical discarded tissues that are subcultured twice before calculation of the PD.

SA-β-Gal activity detection

SA-β-Gal activity was detected as described by Debacq-Chainiaux [42]. Blue staining was detected under an IX71 microscope equipped with DP20 camera and cell^∧A software (Olympus, Hamburg, Germany).

Irradiation

When indicated, exponentially growing HMECs were exposed to 1 Gy of γ-rays using an IBL-437C R-137 Cs irradiator (dose rate of 5.10 Gy/min).

Immunodetection

Flow cytometric analysis

After trypsinization, HMECs were blocked for 20 minutes in PBS-1% BSA, incubated for 30 minutes with anti-CD227-FITC (clone HMPV, Becton Dickinson, Franklin Lakes, NJ, USA) and anti-CD10-PE (clone HI10a, BioLegend, San Diego, CA, USA) at a final concentration of 1:100, all on ice. Flow cytometric analysis was performed using a FACSCanto (Becton Dickinson).

Immunofluorescence

To detect γH2AX and pericentrin, HMECs were fixed in 4% paraformaldehyde for 15 minutes and permeabilized in a 1xPBS-0.5% Triton- X100 for 20 minutes. To detect γH2AX and claudin-4, cells were fixed with ice-cold methanol for 10 minutes. Cells were incubated for 1 hour with blocking solution (1xPBS-0.1% Tween20-3% FBS) before applying primary antibodies mouse anti-γH2AX (Ser139) (clone JBW301, Millipore, Madrid, Spain), rabbit anti-pericentrin (Abcam, Cambridge, UK) or rabbit anti-claudin-4 (Abcam) at 1:1000, 1:2000 and 1:250 final concentrations respectively. Secondary antibodies anti-mouse Cy3 (Jackson ImmunoResearch Inc., Cambridge, UK) and anti-rabbit A488 (Thermo Fisher Scientific, Waltham, MA, USA) were applied at a final concentration of 1:800 and 1:500 (claudin-4) or 1:1000 (pericentrin) respectively. Nuclei were counterstained with 4’,6-diamidino-2-phenylindole (DAPI) at a final concentration of 0.25 μg/ml. For image acquisition an Olympus BX61 epifluorescent microscope equipped with a CV-M4+CL camera (JAI, Grosswallstadt, Germany) and Cytovision software (Applied Imaging, Newcastle, UK) were used.

Automated microscopy and γH2AX foci counting

γH2AX foci counting was done following a semi-automatic approach. Images from slides with γH2AX and pericentrin immunofluorescence were captured using an Olympus BX61 epifluorescence microscope equipped with an automatic motorized stage (BX-UCB, Olympus) and a CCD camera (CV-M4+CL, JAI). The capture methodology was adapted from the Spot-counting system (Spot AX software, Applied Imaging) as described by Hernández [24]. Images were acquired automatically with a 60x objective using predefined settings. Four z-stacks were acquired for γH2AX and 6 for pericentrin, with a step size of 1.55 μm between planes. Cells with only one pericentrin signal were selected and γH2AX foci were scored using FociPicker3D algorithm for Fiji software [43].

Statistical analysis and data modelling

Descriptive analysis and graphics were performed using Microsoft Excel (Microsoft® Excel® 2011, v14.1, Redmond, Washington, USA) and GraphPad Prism 6 (GraphPad Software Inc., San Diego, CA, USA) with methods indicated in the results where applicable. When comparing the number of γH2AX foci/cell among individual donors, Kruskal-Wallis test with Dunn’s multiple comparisons correction was applied and different letters indicate statistical differences (p-value < 0.05) between donors in the graphical representation.

In order to statistically compare the two age groups at each time point analyzed, a generalized linear model with a Negative Binomial distribution response and with repeated measures for each donor was established. The estimated values for γH2AX foci number and the corresponding confidence intervals were obtained using SAS software (SAS v9.4, SAS Institute Inc., Cary, NC, USA).

For the hierarchical cluster analysis, standardized values of γH2AX foci from the 12 donors along the four time points (non-irradiated, 1h, 2h and 24h pIR) were used. Standardized data was obtained by subtracting the mean number of each condition (donor, time and replicate) to the number of γH2AX foci scored for each cell and then dividing this value by the standard deviation of the condition. As it was defined by Everitt and colleagues [44], a hierarchical classification consists of a series of partitions, which may run from a single cluster containing all individuals, to n clusters each containing a single individual. In our case we wanted to determine the inter-group (young vs old) proximity, and thus the Ward method [45] was applied using R software (version 3.4.4, Vienna, Austria). In this method, the criterion for choosing the pair of clusters to merge at each step is based on the size of the error sum-of-squares. Hierarchical clustering is represented by a two-dimensional diagram known as a dendrogram, which illustrates the fusions or divisions made at each stage of the analysis.

A first order kinetic reaction was established to obtain estimations regarding the kinetics of DSB repair in YDs and ADs. This approach was done using methodologies for Nonlinear Regression Model using SAS software (SAS v9.4, SAS Institute Inc.). The established model is described by the following equation:

${\left[n\_foci\right]}_t=\left\{\begin{array}{c}{\theta \mathrm{*}e}^{\beta 1\mathrm{*}t}+{\left[n\_foci\right]}_0 \text{i}\text{f}\text{Y}\text{D}\\ \theta *e^{\beta 1\mathrm{*}t+\beta 0}+{\left[n\_foci\right]}_0 \text{i}\text{f}\text{A}\text{D}\end{array}\right)$

where [n_foci]_t is the number of γH2AX foci at a concrete time after irradiation, 𝜃 is the number of radiation-induced DSBs, β₁ is the γH2AX foci decay proportion, [n_foci]₀ is the basal frequency of γH2AX foci (before irradiation) and β₀ is a constant of delayed repair onset. This model assumes that (1) the same number of DSBs per unit of radiation are induced in YD and AD cells immediately after irradiation 𝜃; (2) cells do not reach complete repair, but instead they reach the basal frequency [n_foci]₀ of DSBs and (3) cells from ADs suffer a delay in DSBs repair initiation (β₀).